AI News Digest 2026-10-24

台本で使った記事

参考記事一覧

参考記事一覧を表示

Donating Agent Skills to the Agentic AI Foundation (importance: 0, dev: 60)
Chinese AI model Kimi K3.1 Pro is closing in on Anthropic's Claude Opus 5.5 (importance: 50, dev: 40)
DeepSeek wrapping up $15 billion fundraising to fuel AI race, sources say - Central Oregon Daily (importance: 60, dev: 10)
WEEK IN REVIEW: U.S. Bans DeepSeek AI From Government Agencies - The China-Global South Project (importance: 40, dev: 10)
DeepSeek, MiniMax secure licences to deploy AI overseas under China export rules - South China Morning Post (importance: 40, dev: 5)
OpenCode AI Coding Tool Exposes API Keys in Suspected Distillation Attack, Triggering Unauthorized Charges - CPO Magazine (importance: 50, dev: 80)
SpaceX Works With Apollo, Nvidia on New Grok Partnership Program - Deccan Chronicle (importance: 40, dev: 30)
CyberCartel hacking gang used Claude and Grok AI agents to automate attacks on Latin American institutions — passwords and details stolen from over 400 institutions, including banks and payment apps - Tom's Hardware (importance: 55, dev: 25)
Lawsuit: Bloodworth created illegal images from photos of student at Walnut Farm Montessori School in Bentonville - Northwest Arkansas Democrat-Gazette (importance: 15, dev: 0)
Judge denies Trump administration request to dismiss NAACP lawsuit against xAI - Northeast Mississippi Daily Journal (importance: 30, dev: 0)
Anthropic、先回りして動くClaude Duetの研究プレビューを公開 (importance: 65, dev: 40)
Canva launches Craft, its first in-house image model (importance: 35, dev: 15)
1Password launches a way for AI agents to log in on your behalf without seeing your password (importance: 45, dev: 50)
Google's Agent Compute Engine gives every Gemini agent its own hardware-isolated micro virtual machine (importance: 60, dev: 80)
Cohere Releases Rerank 5 With Passage-Level Context Focus and Calibrated Scores (importance: 40, dev: 80)
OpenAI、分類やルーティング向けのDecisions APIを一般提供　GPT-6 Lunaベースで最大10倍速、1リクエスト400問とプロンプトキャッシュに対応 (importance: 60, dev: 80)
By every measure, Opus 4.8 beats human experts as peer reviewers in a study of 124 Harvard social science papers covering a range from statistical analyses to formal models to qualitative work. (importance: 30, dev: 10)
what are some everyday examples of AI gadgets (importance: 5, dev: 0)
Major US banks are locking out AI agents that log in with customer credentials (importance: 50, dev: 15)
Security Update on PHP2Shell: Impact for Docker Customers (importance: 40, dev: 75)
Railway、環境変数を保管するシステムに不正アクセス　一部は復号済みの値が取得された可能性、全ユーザーにシークレットのローテーションを推奨 (importance: 60, dev: 90)
OpenAI lowers valuation target, raises up to $30 billion (importance: 65, dev: 10)
Uber brings ChatGPT Work to all employees with OpenAI (importance: 35, dev: 15)
Meta’s AI could still scrape news sites, legal experts say (importance: 30, dev: 5)
Three-week on-site audit at OpenAI reveals major gaps in AI safety monitoring (importance: 45, dev: 15)
Today, we hacked what we'd set out to hack: the iPhone (importance: 30, dev: 25)
Job loss and cortisone mania (importance: 2, dev: 0)
Docker is weird (importance: 30, dev: 85)
Local Wikipedia: Wikipedia, but only for things a half hour walk from you (importance: 10, dev: 25)
LLMs are making formal software verification feasible at a much larger scale (importance: 55, dev: 80)
Canadians are running out of time to oppose Bill C-22 (importance: 25, dev: 5)
How does Fable 5.1 hide messages in images? (importance: 40, dev: 10)
Ask HN: If the models are so good, why doesn't OpenAI just start 100 companies? (importance: 15, dev: 10)
OpenShell – A safe, private runtime for fleets of autonomous AI agents (importance: 55, dev: 85)
Show HN: OnKey – I'm a bad singer, so I built an app to show where I'm off-key (importance: 10, dev: 30)
The alien space of science: Coherent but cognitively unavailable directions (importance: 30, dev: 10)
Open call for guest investigators in the life sciences (importance: 20, dev: 5)
v2.1.309 (importance: 45, dev: 90)
v2.1.308 (importance: 10, dev: 80)
v1.19.4 (importance: 30, dev: 85)
Rapidly scaling online storage to serve over 1 billion ChatGPT users (Part II) (importance: 40, dev: 70)
OpenAI and Poland’s NASK partner to advance AI safety and security (importance: 20, dev: 10)
What FDE means at OpenAI (importance: 25, dev: 30)
Bringing foundational models to the front lines of TB screening (importance: 20, dev: 10)
Flint-interactive: A language for creating interactive data visualizations (importance: 35, dev: 60)
LitPivot: Developing research ideas through dynamic contextualization within the literature landscape (importance: 30, dev: 35)
Does Your Model Run on AWS Trainium? Look It Up (importance: 30, dev: 60)
Turning Database Grading into RL Rewards: Post-Training Agents with ThinkingBox (importance: 35, dev: 60)
Ensemble forecasting with IBM Granite time-series models (importance: 20, dev: 40)
How bug bounty sharpens one researcher’s security skills (importance: 15, dev: 40)
Origin Range Requests: Serving large files without range-request tuning (importance: 35, dev: 85)
Bringing post-quantum cryptography to geographically fenced key management (importance: 30, dev: 60)
Localizing compliance at scale: achieving South Korea’s ISMS certification (importance: 15, dev: 20)
JetBrains GameDev Day 2026: Watch the Recordings (importance: 10, dev: 40)
Air Governance: Manage AI Agents, Access, and Spend Throughout Your Organization (importance: 0, dev: 70)
ActiveState Partners with Docker to Secure the Open Source in Enterprise Containers (importance: 15, dev: 50)
Scaling NVIDIA TensorRT Performance Optimization Beyond a Single Build Route with the Global Performance Tuner, Part 2 (importance: 35, dev: 85)
Scaling NVIDIA TensorRT Performance Optimization Beyond a Single Build Route with the Global Performance Tuner, Part 1 (importance: 35, dev: 85)
Using AI Agents to Teach Robots Like Microduck to Walk in Simulation (importance: 25, dev: 40)
How to Read Data into GPU Memory in Python with NVIDIA GPUDirect Storage (importance: 25, dev: 85)
Scaling Decision Optimization to 100 Million Variables and Beyond with mPDLP in NVIDIA cuOpt (importance: 30, dev: 75)
Designing an AI serving harness built to last (importance: 30, dev: 55)
The Foundations of Scalable AI (importance: 10, dev: 15)
Making the case for context-aware AI (importance: 10, dev: 20)
AI’s hidden storage bill: What inference leaves behind (importance: 20, dev: 30)
Intel gets waiver after worries fuzzy US ban on foreign-made robots may hit fabs (importance: 15, dev: 0)
Uber rival inDrive eyes 2027 robotaxi tests in Central Asia with Pony AI and Nvidia (importance: 15, dev: 5)
Left alone with nothing to do, AI agents started playing (importance: 35, dev: 25)
OpenAI makes its Jev competitor widely available as decision model competition heats up (importance: 55, dev: 70)
Microsoft’s Mustafa Suleyman says its AI outperforms doctors on hard cases as he chases ‘medical superintelligence’ (importance: 25, dev: 5)
Chinese hackers spied on thousands of companies after breaching IT giant ConnectWise (importance: 55, dev: 35)
Spotify’s current AI bet is about being the music service that agents choose (importance: 25, dev: 15)
Nirva raises $33M for a pendant that only listens to you (importance: 10, dev: 5)
The iPhone Duo should have been an iPad (importance: 10, dev: 0)
Now Google says it might use your Gmail and Google Photos data to target ads (importance: 35, dev: 5)
Retro gaming is being overrun with AI (importance: 15, dev: 5)
Meta turns Muse sharing back on (importance: 30, dev: 5)
Sure, let’s smoosh SpaceX and Tesla together, who cares (importance: 5, dev: 0)
Discord now formats Markdown previews right (importance: 15, dev: 25)
In Meta’s version of the future, I’ll have my Muse call your Muse (importance: 15, dev: 5)
Anthropic Adds Opt-In Security Scanning for Critical Open-Source Projects (importance: 65, dev: 75)
Study Finds AI-generated Code Fails to Boost Productivity Due to Downstream Bottlenecks (importance: 0, dev: 70)
Presentation: Actors for Agents: Patterns for Production AI (importance: 35, dev: 80)
Vercel Labs Open-Sources fx: A Zig Coding Agent Designed to Be Embedded (importance: 50, dev: 85)
Article: AI Changes the Data Interface, Not the Query Engine (importance: 35, dev: 70)
Developer Behind the AI-Built Adobe Clones Says They Are “Not Anywhere Close to Ready (importance: 35, dev: 40)
Why Engineers Should Brave Capitol Hill (importance: 15, dev: 20)
Anthropic says its unreleased model is both more capable and better aligned than Mythos 5.1 (importance: 70, dev: 40)
OpenAI CEO Sam Altman says advanced AI will inevitably escape its sandbox, and only alignment can save us (importance: 35, dev: 10)
Anthropic's Claude now writes code and proves it safe automatically (importance: 55, dev: 80)
Only 13 percent of companies see real returns from AI, Gartner survey finds (importance: 30, dev: 5)
Strands AgentsのA2Aサーバ、会話状態が利用者間で漏れる欠陥 (importance: 55, dev: 85)
AWS SAM CLIのTerraformフックにコマンド注入、1.166.2で修正 (importance: 40, dev: 90)
GitLab、正規表現コンパイラの欠陥で未認証の任意コード実行を修正 (importance: 60, dev: 80)
Zod作者のColin McDonnell氏、.sqlをTypeScriptに直接importできるReyloを公開　TS 7.1のcontent mappersで型を推論、コード生成なし (importance: 35, dev: 90)
Linuxのksmbdに未修正のCVSS 9.0脆弱性2件　ZDIの報告が誤って公開メーリングリストに投稿され、再現手順まで流出 (importance: 40, dev: 50)
HTTP/3のQPACKを悪用するDoS「Header Accordions」　FoxIOが10実装中7つで確認、quic-goやquicheは修正済み、Bunやh3は未修正 (importance: 40, dev: 60)
BIND 9にCVSS 8.1の脆弱性CVE-2026-94300　GSS-TSIG署名済みUDPメッセージの再送でメモリ破壊、9.20.30と9.21.27で修正 (importance: 40, dev: 60)
PostgreSQL 19のRC1が公開、正式版は11月19日の予定 (importance: 35, dev: 70)
Cloudflare、要求と応答を組で判定するWAFを全プランで提供 (importance: 40, dev: 70)
GitHub、npmの取得を検査するパッケージファイアウォールを公開 (importance: 0, dev: 75)
Vercel、Pythonでルーティングミドルウェアを書けるように (importance: 0, dev: 80)
Chrome 156が安定版に、重大6件を含む353件の脆弱性を修正 (importance: 50, dev: 40)
OpenAI、Responses APIに12時間のプロンプトキャッシュを追加 (importance: 0, dev: 80)
Google Cloud、OpenAIのGPT-6系をModel Gardenで限定公開 (importance: 62, dev: 60)
TinyFish Breaks Down Agentic Search Architecture for Production AI Agents (importance: 35, dev: 80)
You.com Launches Inference Router: 12 Open-Weight Models With 99.9% Uptime SLA (importance: 45, dev: 70)
Liquid AI Launches RIPL, a Real-Time Inference Engine for Sub-50 ms LFMs (importance: 35, dev: 65)
Gradium Open-Sources Phonon, a 31M-Parameter On-Device TTS Model Under Apache 2.0 (importance: 35, dev: 45)
A Developer’s Guide to OmniParseBench: Leaderboard Benchmarks and OCR Testing (importance: 15, dev: 75)
IBM AI Releases Granite Speech 5.0 2B for Multilingual Transcription and Translation (importance: 35, dev: 35)
Perplexity and NVIDIA Researchers Introduce JOLT: a Joint Teacher-Student Reinforcement Learning Method (importance: 40, dev: 40)
A Coding Guide to Cactus Whistle: Does a 16.9 MB Speech Model Hold Up Against Whisper Base? Word Error Rate, Speed, Streaming, Timestamps, Keywords, and Speech-to-Tool Calls on a CPU (importance: 25, dev: 70)
The Cost of Saying Hello: An Empirical Study of Handshake Overhead in AI Agent Communication Protocols (importance: 40, dev: 70)
MCP vs CLI for AI Agents: A Comparative Analysis of Tool Integration Strategies for Autonomous Systems (importance: 50, dev: 80)
Accelerating Wildfire Hazard Mapping Across Canada with an AI Surrogate Modelling Framework (importance: 15, dev: 10)
Neural Networks with Differentiable Belief-Function Layers (importance: 25, dev: 30)
When Stable and Preferred Semantics Agree: A Path-Based Approach (importance: 10, dev: 10)
Feedback That Backfires in Games: Do LLM Players Repeat the Move They Just Watched Fail? (importance: 35, dev: 40)
Breaking the high-degree barrier in Lehmer's problem with a human-AI discovery loop (importance: 30, dev: 5)
Route-and-Generate: KL-Regularized Router Improvement for Reasoning Routing in LLMs (importance: 25, dev: 30)
Soft Ontology Priors over Latent Concept Posteriors for Weakly Supervised Neuro-Symbolic Learning (importance: 15, dev: 10)
RoleFlow: Self-Evolving Multi-Role Agentic Collaboration via Structural Adaptation (importance: 25, dev: 45)
Secure Agentic AI and Multimodal Intelligence in 6G Communication Systems (importance: 15, dev: 15)
SpecMAGIC: Specification-Guided Multi-Task Generalization in Multi-Agent Reinforcement Learning (importance: 15, dev: 15)
A Three-Pillar Governance Framework for Safe and Trustworthy AI Agents (importance: 20, dev: 30)
Same Beliefs, Different Policies: Learning Reusable Representations for Early Stopping in LLM Reasoning (importance: 25, dev: 35)
Why Bother Me Again? Benchmarking Personal Agent Memory for Interaction Preferences (importance: 25, dev: 35)
From Single-Actor Environments to Multi-Actor Event Worlds: Can LLMs Be Event World Models? (importance: 15, dev: 20)
What the Trace Shows and What the Label Says: Auditing Behavioral Detectors in Agent Evaluation (importance: 30, dev: 60)
From Prose to Properties: Testing Research Claims with ARATest for Agent-Native Science (importance: 20, dev: 50)
Can LLMs Improve Their Own Harness? A Controlled Study on ARC-AGI-3 (importance: 35, dev: 60)
PLE: Physical-Engineering's Last Exam (importance: 15, dev: 10)
Governed Self-Evolution for Autonomous Investing: A Multi-Agent LLM System with an Invariant Core (importance: 15, dev: 20)
Diagnosing Personalization Failures in Mobile GUI Agents (importance: 20, dev: 35)
Model Organisms of Collusion: Corrective Transfer in Untrusted Monitors (importance: 25, dev: 25)
Better Reasoning, Not Just Bigger Models: A Structure-Grounded Framework for Efficient and Reliable Inference (importance: 15, dev: 30)
Large language models show systematic failures in incorporating evolving medical knowledge (importance: 15, dev: 5)
Speculative Acceptance Does Not Decay with Position: A Full-Length Diagnostic on Long Chain-of-Thought with Independent Drafts (importance: 30, dev: 70)
Anticipating Social Consequences with Distributional World Models (importance: 15, dev: 15)
A Topology-Guided GeoAI Agentic Framework for Urban Flood Screening (importance: 10, dev: 20)
Group Relative Policy Optimisation Improves Multilingual Speech Recognition in Low-Resource Languages (importance: 20, dev: 20)
Scaling Laws for Safety Probes (importance: 30, dev: 40)
Disentangling Robustness and Sycophancy in Multilingual LLMs (importance: 20, dev: 15)
Governance-State Corruption in Long-Term Memory for LLM Agents: A Cross-Model Stress Test (importance: 35, dev: 60)
Faithful Latent Harm Monitors Work No Better Than the Models Own Behaviour at a Matched False Alarm Rate (importance: 25, dev: 20)
AgentTriage: Replayable Security Diagnosis Across Heterogeneous Agent Runtimes (importance: 30, dev: 65)
How Firm Are AI Models' Ethical Boundaries? A Pay-Stub Editing Test of 15 Models (importance: 35, dev: 10)
Is Complex Agentic Memory Always Needed? Naive RAG Is Stronger Than We Thought (importance: 30, dev: 60)
Can Loan Processors and Underwriters Spot AI-Edited Statements? An Empirical Study (importance: 15, dev: 5)
AgentDP: Towards Automated Model Deployment for Heterogeneous Embedded AI Applications (importance: 15, dev: 40)
Scaling Laws for Generative Reward Models (importance: 25, dev: 30)
LLM Multi-Agent Systems Are Vulnerable to Herding and Runaway Cascades (importance: 25, dev: 30)
ProQuant: Preserving Expert Quality over Quantity in Ultra-Low-Bit Quantization of Mixture-of-Experts Models (importance: 25, dev: 50)
An Introduction to World Models for Generalist Agents (importance: 20, dev: 25)
Agentic Knowledge-Model Co-Evolution via Consistency Regularization (importance: 25, dev: 40)
What Lies Beyond Perception: A Partition Theory of the Limits of AI Evaluation and Oversight (importance: 20, dev: 20)
TESSA: A Governance-Oriented Custom GPT for SME Guidance under the EU AI Act: Design, Implementation, and Evaluation (importance: 10, dev: 10)
Semi-Linear Gaussian Bayesian Networks: Enhancing Flexibility in Node Dependencies via penalized B-Splines (importance: 5, dev: 10)
ADAPT for Science: A Multi-Agent System for Self-Evolving Scientific Optimization (importance: 20, dev: 30)
Parameter Learning in Active Inference via Variational Message Passing (importance: 10, dev: 10)
Beyond Phenotypic Similarity: Relation Refinement and Selective Population Graph Learning for Multimodal Brain Disorder Diagnosis (importance: 10, dev: 5)
TwinStore: A Synthetic Consumer Platform Using Vision--Language Models (importance: 20, dev: 20)
Harnessing Language Models as Optimization Modeling Agents (importance: 30, dev: 40)
DConstrain: Compiling Context-Parameterized Transitions for Diffusion Language Model Constrained Decoding (importance: 30, dev: 45)
The Geometry of Meaning in Neural Manifolds (importance: 15, dev: 10)
Measuring the Reasoning-Action Coupling in Driving Vision-Language-Action Models through Counterfactuals (importance: 15, dev: 10)
Zero-Shot Forecasting of In-Situ Soil Moisture and Temperature with a Time-Series Foundation Model (importance: 10, dev: 10)
ColaBench: Evaluating Memory-Guided Knowledge Acquisition in Personal Agents (importance: 20, dev: 25)
Agri-SCHOLAR: Simulation-Grounded In-Season Rolling Advisories for Agricultural Crops Using LLM-Bounded Optimization (importance: 10, dev: 15)
Mitigating Saliency Collapse: Robust Saliency-Aware Long-Text Image-Text Alignment (importance: 20, dev: 20)
Evaluating Artificial Wisdom (AW): A Trajectory-Based Framework for Context-Sensitive Decision Regulation (importance: 10, dev: 15)
CANNBot-CAKE: Collaborative Agentic Kernel Evolution for Ascend NPU (importance: 25, dev: 60)
Multichoice Modeling: Exploring Decision Configuration Space under Uncertainty (importance: 10, dev: 10)
Algorithmic drift in generative adaptive learning (importance: 10, dev: 5)
Conflict-Aware Logit Adapters for Utility-Preserving Anti-Distillation (importance: 25, dev: 35)
Directing to Diversify: Multi-Directional Search for Automatic Heuristic Design (importance: 20, dev: 30)
Diagnosing the Explicit-to-Applied Theory of Mind Gap in LLMs (importance: 15, dev: 10)
SCOPE: Compiling Natural-Language Policies into Neurosymbolic Predicate Programs (importance: 30, dev: 55)
Dual Exam: Controlled Evaluation of Multimodal Agents and World Models (importance: 20, dev: 30)
Co-Recursive Semantic Self-Extension: A Mathematical Foundation for Open-Ended Intelligence (importance: 5, dev: 5)
Solving Flood-It Puzzles: A Comparison of Tailored A* Search with Declarative Solver Approaches (importance: 15, dev: 40)
DeepOpen: Decision-Oriented Intent Classification on BANKING77 and CLINC150 (importance: 15, dev: 30)
Who Is the Influencer? Consensus Dynamics in Multi-Agent Systems (importance: 20, dev: 15)
Can AI Co-Design Structures? An AI Agent for Early-Stage Structural Design (importance: 15, dev: 15)
Constraint Reasoning and Large Language Models Join Forces: Providing Contrastive Explanations for Real-World Optimisation-Based Applications (importance: 15, dev: 25)
SetCommit: Coordinated Token Commitments for Few-Step Diffusion Language Model Decoding (importance: 20, dev: 30)
Post-Audit Mirage: Identifiability Limits of Offline Agent Verification (importance: 20, dev: 30)
Rethinking Fisher Vectors for Second-order Multi-scale Deep Feature Aggregation in Scalable Scene Retrieval (importance: 10, dev: 15)
Different Tasks, Different Timesteps: Temporal Task-Aware Quantization for Diffusion Language Models (importance: 15, dev: 25)
Trion-1: Unifying Language, Protein Sequence, and Structure with Diffusion Language Models (importance: 15, dev: 5)
Reward-Weighted Preference Optimization for Selective Combinatorial Optimization Problems (importance: 10, dev: 15)
Reducing Multi-Horizon Photovoltaic Forecast Error through Tree-Anchored Expert Combination (importance: 10, dev: 15)
PointerBench: A Diagnostic Benchmark for Fine-Grained GUI Grounding (importance: 20, dev: 35)
RIPPLE: Recovery Internalization through Process-Level Policy Learning and Effect-Aware Credit for Long-Horizon Agentic Coding (importance: 30, dev: 70)
The Value of Memory in Language Agents (importance: 20, dev: 30)
Subgoal-aware Memory Graph for GUI Agent (importance: 15, dev: 25)
ValuSpec: Plug-and-Play Candidate Valuation before Target Verification for Tree-Based Speculative Decoding (importance: 20, dev: 50)
Muffin - A Shared Memory Buffer for Asynchronous AI Agents (importance: 20, dev: 40)
RIDAR: Safeguarding Research Integrity in Autonomous Research Agents via Automated Detection and Runtime Intervention (importance: 25, dev: 30)
Certified Execution Harness for Agentic AI: Safe and Least-Intrusive Petri-Net Supervisory Control Between Reasoning and Action (importance: 45, dev: 65)
PhysTrace: Interpretable Temporal Evaluation of Physical Consistency in Video Generative Models (importance: 25, dev: 15)
Conservative Meta-Agent Search: Architecture Adaptation under Non-Stationarity (importance: 30, dev: 45)
DIAL: Steerable Language Models with Distilled Instruction-Adherence Levels (importance: 30, dev: 35)
Adaptive Skill Granularity for Coding Agents: Runtime Routing and Macro-to-Micro Recovery (importance: 40, dev: 70)
FairFRESH: A Neuro-Symbolic Framework for Fair and Explainable Household Food Recommendation (importance: 10, dev: 10)
DuoCook: An Overcooked-Inspired Benchmark for Visual Planning and Coordination (importance: 15, dev: 20)
From Prompt Rules to Golden Data Model: Production-Scale Generation of Personalized, Compliant Pharmaceutical Claims (importance: 25, dev: 30)
Jev-DLC: Decision-Level Continual Learning from Typed Feedback (importance: 20, dev: 20)
Beyond Causal Masks: Auditing and Localizing Hidden Future-Token Leakage in Adaptive Sequence Models (importance: 30, dev: 35)
Contestable Agents for Data Governance: Recording Human Stewardship of AI-Generated Metadata Across Structured and Unstructured Systems (importance: 15, dev: 20)
What Makes a Good Unmasking Policy? A Controlled Study of Learned Unmasking in Diffusion Language Models (importance: 20, dev: 10)
Who Is Sycophantic to Whom? Confident LLMs, Human Deference, and the Collapse of Trust Calibration (importance: 25, dev: 10)
Simulated Control Evals (importance: 25, dev: 15)
EpiSEM-Bench: Evaluating Safety and Evidence Recovery in LLM Agents under Missing Evidence (importance: 30, dev: 45)
Direct Pass@$k$ Optimization for Thought Pruning: Balancing Quality and Diversity in Test-Time Reasoning (importance: 20, dev: 20)
Lighthouse: On Building a Trustworthy Digital Assistant for Exploring History with Civil Rights Archives (importance: 15, dev: 10)
V-JEPA 2 Meets AQA: Phase-Based Scoring and Reference-Based Feedback for Mountain Biking (importance: 10, dev: 5)
TauIndianBankBench: Procedure-Aware, Action-Oriented Evaluation of Tool-Using Conversational Agents (importance: 25, dev: 35)
FusionEval: Benchmarking Model Fusion for Large Language Models (importance: 25, dev: 30)
IRAH: Hierarchical Agentic Tree Search for Retrieval Pipeline Optimization (importance: 25, dev: 40)
GeoDiv KAN Solver for Time Fractional Navier Stokes Equations on Complex Domains to Realistic CFD Reconstruction (importance: 8, dev: 5)
Detecting AI Training in Datacentre Power (importance: 30, dev: 5)
Self-Evolving Static Analysis via LLM Pseudo-Labeling (importance: 35, dev: 70)
Agentic Verified Optimization of Real-World Software (importance: 40, dev: 75)
Why Constrained Updates Still Forget: A Functional-Path Analysis of Continual SFT (importance: 25, dev: 20)
Detection Is Not Enough: Reasoning Traces as Testable Program Hypotheses (importance: 25, dev: 30)
MPIDetector: Detecting Multi-modal Prompt Injection via LLM-based Reasoning (importance: 25, dev: 25)
MoEZip: Routing-Aware KV Cache Compression for Sparse Mixture-of-Experts LLMs (importance: 25, dev: 30)
Prediction Is Not Calibration: Item Parameter Recovery and Representation Coupling in Knowledge Tracing (importance: 10, dev: 5)
When Reconstruction Is Not Enough: Doubly Robust Counterfactual and Interventional Mediation Analysis with Learned Multimodal Representations (importance: 8, dev: 3)
Automated screening of privacy risks in medical AI (importance: 15, dev: 15)
Edit, Don't Imitate: Minimal-Edit On-Policy Distillation for Multimodal Reasoning (importance: 20, dev: 10)
SWE-VERIFIER: Selective Verification for Cost-Efficient Coding Agents (importance: 40, dev: 70)
Generalizing Metropolis-Hastings for DAG Sampling (importance: 8, dev: 5)
GAV: A scientific experimentation environment for PDE agents (importance: 20, dev: 35)
Gender, Sexual Orientation, and the One-Year Refugee Bar: A Computational Study of Bill C-12 and Canadian Refugee Adjudication (importance: 10, dev: 3)
TerraTransfer++: Demonstration-Free End-to-End Planner (importance: 20, dev: 10)
MPBRS: Multi-Potential-Based Reward Shaping for Unified Reward Configuration in Humanoid Locomotion (importance: 15, dev: 15)
Consistency Is Not Validity: Self-Rewarded Moral Argumentation Makes Language Models Decisive but Pliable (importance: 20, dev: 10)
Probing and Steering for Safety-Relevant Behaviours in Large Language Models (importance: 30, dev: 15)
MemPlan: Memory-Conditioned PDDL Planning for Partially Observable Text Environments (importance: 20, dev: 25)
AcoustiPrint: Physics-Guided Neural Design of Manufacturable Acoustic Structures (importance: 8, dev: 3)
Agent Mosaic: Recognizing Confidentiality Leaks in Agentic LLM Systems (importance: 35, dev: 40)
High-Value Training Data from Noisy, Diverse Sources (importance: 25, dev: 20)
Self-Assessment Reinforcement Learning for Search-State Awareness in Retrieval-Augmented Large Language Models (importance: 25, dev: 30)
aggmetrics: Aggregating Classification Performance Measures for Machine Learning (importance: 15, dev: 45)
Short-to-Long: Detached Features for Long-Horizon Value Learning (importance: 15, dev: 5)
SplatLab: Autonomous Scientific Discovery for 3DGS Research (importance: 25, dev: 25)
Rethinking SAE Evaluation: An Atomic Interpretable Unit Framework Reveals Hidden Polysemanticity and Redundancy (importance: 20, dev: 5)
Externalise the Memory: A Pre-Registered Study of Where an Embodied Agent's Experience Should Live (importance: 20, dev: 15)
PhysibleMDM: Enforcing Inference-Time Physical Feasibility Constraints on Motion Diffusion Models via Noise Optimization (importance: 10, dev: 5)
Behind Aggregate Scores: What Systematically Varying Arena-Hard's LLM Judge Reveals (importance: 30, dev: 35)
Protegrity-PII Technical Report (importance: 20, dev: 30)
LLM2Cert: LLM Synthesis of Certificates with SMT/Lean Verification (importance: 20, dev: 15)
Decision Making and Ride Comfort Evaluation Based on Lane-Change Prediction for Autonomous Vehicles (importance: 10, dev: 5)
Lifetime-valid decisions by continually evolving AI analysts (importance: 25, dev: 15)
Bridging the Closed-World Gap: Interpretable Anomaly Detection via a Hybrid TSK-FCM Ensemble on the BETH Dataset (importance: 10, dev: 15)
Visual Grounding First, Multimodal In-context Learning Follows (importance: 20, dev: 10)
Impacts of generative AI skill development on academic adaptation and learning outcomes among university students (importance: 15, dev: 5)
Future of Time Series Anomaly Detection: From Historical Review to Unified Outlook (importance: 20, dev: 10)
Interpreting Model Reasoning through Behavioral Counterfactuals (importance: 25, dev: 15)
Asking Less, Adapting Better: Feedback-Efficient Test-Time Adaptation for Vision-and-Language Navigation (importance: 10, dev: 5)
One-shot federated unsupervised domain adaptation with Smoothed Knowledge Distillation and teacher refinement (importance: 10, dev: 10)
Spectral Smoothing and Spiking Perturbations Reconfigure Capability Profiles and Selective Task Performance in Large Language Models (importance: 20, dev: 10)
From Clinical Judgment to Checkable Rules: Clinician-Led Guardrails for Mental-Health Chatbots (importance: 20, dev: 25)
LoRA Superposition across Pretrained Models (importance: 30, dev: 25)
ULME: Predictive Coupling for Mutually Reinforcing Embedding and Language Modeling (importance: 25, dev: 20)
Meaning Contracts: Local Semantic Geometry in Large Language Models (importance: 15, dev: 5)
Learning from Next-Frame Prediction: Towards Unified and Scalable Visual Representation Learning (importance: 20, dev: 5)
The Delegation Divide: An Open Index of AI-Agent Delegability across Occupations and Countries (importance: 35, dev: 20)
Abstract Priorities, Concrete Rewards: How Stated Preferences Relate to What People Reward in AI Responses (importance: 15, dev: 10)
VLTest: Finding Joint Image-Text Failures in Vision-Language Models (importance: 20, dev: 35)
Neural Spectral Capacity with Residual Context for Heterogeneous Resource Allocation within Transformers (importance: 15, dev: 15)
One Evaluation Is Enough: Inversion-Free Image Editing with Trajectory Correction (importance: 12, dev: 5)
Bifurcations from Weight Space: JEPAWG on Parametric Dynamical Systems (importance: 8, dev: 3)
Direct Forcing: A Single-Stage Training Recipe for Few-Step Autoregressive Text-to-Video Generation (importance: 15, dev: 5)
Separate Field Scoring Protects Unused Tools When Agents Pool Tool-Use Experience (importance: 25, dev: 45)
MOER: Mix-Policy Optimization under Expert Prefix Guidance with Selective Rollback (importance: 15, dev: 10)
Query-Time Calibration of CLIP Features for Open-Vocabulary 3D Segmentation (importance: 10, dev: 5)
MIG-COW: Transferable Adversarial Attacks on Deepfake Detectors via Gradient Decomposition (importance: 10, dev: 5)
Low-Latency MoE Inference in Space Data Centers: A Wireless-Power-Transfer Perspective (importance: 12, dev: 10)
Vision Transformers for Estimating Irradiance Using Data Scarce Sky Images (importance: 8, dev: 3)
GateFormer: Forecast-Space Decomposition for Intra-Hour Solar Irradiance Prediction from Sky Images (importance: 8, dev: 3)
Metacognitive Scaffolds Create Productive Friction for AI-Assisted Knowledge Work (importance: 25, dev: 15)
RetiJev: Fast and Source-Grounded Ophthalmic Decision Models (importance: 12, dev: 15)
OpenAttention: One Percent Is Enough with Tail Proxies (importance: 30, dev: 40)
LAST-V2X: Language-Driven Adaptive Scenario Synthesis and Testing for Communication-Induced Failures in V2X Safety Applications (importance: 10, dev: 15)
Truthful by Omission: Click Thinning for LLM-Native Advertising (importance: 25, dev: 10)
SecLog-Bench: A Scalable Benchmark for Cybersecurity Log Reasoning with Scaffolded Evidence (importance: 20, dev: 20)
Multi-Tool Selection for Robotic Systems: Comparing Semantic Retrieval and LLM (importance: 20, dev: 25)
Mostly Reading: What Open-Book Scores Measure in Document-Level LoRA Adapters (importance: 25, dev: 15)
Harm Is Behavioural, Moderation Is Lexical: How Language and Culture Decide Whom Automated Safety Systems Protect and Whom They Silence (importance: 20, dev: 10)
MQAScore: Fine-Grained Text-Music Alignment Evaluation via Large Audio Language Model (importance: 8, dev: 5)
What Pairwise Evaluation Misses: Reward Scores across Prompts (importance: 20, dev: 20)
A Grammar of French Civil Cassation: A Ground-Level Dataset and Tool (importance: 8, dev: 5)
ProductWebBench: Evaluating Website Continuity Beyond Task Completion (importance: 35, dev: 60)
Mixing Matters: Evidence Position Bias Across Sequence Mixers in Long-Context Question Answering (importance: 25, dev: 15)
PriMa-Causa: Causal Prescriptive Maintenance (importance: 12, dev: 10)
Prompt Sensitivity in Tool-Augmented Model Comparisons (importance: 25, dev: 25)
Network topology shapes consensus and vulnerability in LLM populations (importance: 20, dev: 15)
Prompting Strategy Effects Depend on the Judge: A Multilingual Study of LLM-as-a-Judge Robustness (importance: 30, dev: 30)
When More Metadata Misleads: Probing Field-Level Grounding in LLM-Generated Museum Descriptions (importance: 10, dev: 5)
Predicting Music Valence and Arousal via Textual and Multimodal In-Context Learning in Large Language Models (importance: 10, dev: 5)
QICOM: Open-Source Software for Reproducible Image Quality Evaluation, Blinded Assessment, and Radiomic Comparison (importance: 10, dev: 25)
Can Entanglement-Induced Feature-Map Designs Improve Projected Quantum Kernels? (importance: 10, dev: 5)
Notation3: Graph Terms for RDF (importance: 15, dev: 30)
Memory-Aware Reward Modeling: Learning Human Preferences for LLM Memory Use (importance: 35, dev: 25)
CoWriter: Guiding Methodological and Synergetic Human-LLM Co-Authoring (importance: 25, dev: 15)
Beyond Accuracy: Disentangling Decomposition from Typed-State Quality Assurance in Multi-Agent LLM Pipelines (importance: 35, dev: 50)
RecallGS: Geometry-Grounded Recall for Online Semantic Gaussian Splatting (importance: 8, dev: 5)
Superintelligent Software Engineering: Governance That Scales with Capability (importance: 30, dev: 55)
When LLMs Forget What They Believe: Characterizing State Integrity Collapse in Long-Horizon Dialogues (importance: 35, dev: 35)
SurveyGraph: Generating Academic Surveys via Methods-Entity Knowledge Graphs (importance: 20, dev: 15)
Hyperreduction accelerates finite element model updating for the discovery of neural materials (importance: 8, dev: 5)
Bridging the Data Gap: Source-Informed Task-Space Transport for Few-Shot Diffusion (importance: 12, dev: 5)
AgenticEvolve: Evolutionary Program Search with Persistent Language Model Agents that Design and Run Experiments (importance: 30, dev: 35)
Automating Feature Flag Cleanup at Scale with Multi-Agent LLM Systems: An Industrial Experience at DoorDash (importance: 45, dev: 70)
AI Risk Ontologies for the EU AI Act Compliance: AIRO, VAIR, and their Integration with DPV (importance: 15, dev: 10)
Auditing the Hand-Written Inputs of Learned Access Control (importance: 12, dev: 25)
HERD: Harness-Evolving Recursive Distillation for Continual Learning (importance: 30, dev: 40)
AgentPhishing: A Benchmark of LLM Email Agents Under Adaptive Attack (importance: 25, dev: 20)
Same Entity, Different Script: Diagnosing Cross-Script Entity Matching Failures in Language Models (importance: 15, dev: 5)
Major Revision, Minor Accountability: LLMs as Reviewers of Published Articles in Top-Tier Journals (importance: 15, dev: 5)
Post-Hoc Feature Selection Layer for Neural Networks Interpretability (importance: 12, dev: 10)
How Depth Growth and Looping Shape In-Context Learning in Language Models (importance: 25, dev: 15)
FocusVLA: Future-Intent Guided Dynamic View Routing for Vision-Language-Action Models (importance: 12, dev: 5)
Predictable performance is suboptimal, and what it means for model routing (importance: 30, dev: 40)
MoG-VLM: Efficient Video Anomaly Detection through Motion-Guided Vision-Language Models (importance: 10, dev: 5)
Training Small Language Models for Transferable Tool Selection (importance: 35, dev: 50)
SABLE: Scalable Affine Bayesian Latent-space Exploration (importance: 10, dev: 5)
The Geometry of Metaphor: Probing and Steering Metaphor Representations in Large Language Models (importance: 15, dev: 5)
When to Follow the Teacher: Advantage-Based On-Policy Distillation (importance: 35, dev: 25)
Social-HUMINT: An LLM-Augmented Unified Framework for Measuring Social Media Exposure (importance: 15, dev: 10)
Would an LLM Match into a Residency? Uncertainty, Reasoning and Image Use of Open Language Models on the Spanish MIR Examination (importance: 10, dev: 5)
VERA: A Responsible AI Dashboard for Assessing LLM Deployments in Light of the EU AI Act (importance: 25, dev: 30)
How Ground Truth Shapes Retrieval Scores: Evidence-Aware Evaluation of a Retrieval-Augmented Generation System for Breast Lesions (importance: 10, dev: 20)
Questioning Instrumental Convergence in RLVR: Checkpoint Evidence from OLMo-3 (importance: 20, dev: 15)
Causal Chain of Thought: Mitigating Attention Leakage via Context Isolation (importance: 25, dev: 20)
Low-Bit N-Gram Embeddings for Language Models (importance: 35, dev: 45)
METTA: Memory-Enhanced Test-Time Adaptation for Meta-Reinforcement Learning (importance: 12, dev: 5)
Cost-Aware Abstention for LLM-Based IT Ticket Classification (importance: 25, dev: 35)
Anchoring Distributed Agency: Institutional Metaphors for Designing Large-Scale AI and AI Techniques for Running Institutions (importance: 8, dev: 5)
Risk-Aware Alignment Must Incorporate Reward Uncertainty (importance: 12, dev: 5)
Agent-Environment Alignment via Automated Interface Generation (importance: 25, dev: 40)
Group Contrastive Policy Optimization: Learning Beyond the Self-Exploration Limit with Golden Answers (importance: 15, dev: 10)
How to Teach Large Multimodal Models New Skills (importance: 15, dev: 5)
MENTOR: A Metacognition-Driven Self-Evolution Framework for Uncovering and Mitigating Implicit Domain Risks in LLMs (importance: 15, dev: 5)
Survival is the Only Reward: Sustainable Self-Training Through Environment-Mediated Selection (importance: 12, dev: 10)
GFlowPO: Generative Flow Network as a Language Model Prompt Optimizer (importance: 20, dev: 30)
SleepMaMi: A Universal Sleep Foundation Model for Integrating Macro- and Micro-structures (importance: 8, dev: 5)
Sci-CoE: Co-evolving Scientific Reasoning LLMs via Geometric Consensus with Sparse Supervision (importance: 10, dev: 5)
CLEAR: Context Augmentation from Contrastive Learning of Experience via Agentic Reflection (importance: 25, dev: 40)
SkillFlow: Flow-Driven Recursive Skill Evolution for Agentic Orchestration (importance: 25, dev: 40)
Projecting Latent RL Actions: Towards Generalizable and Scalable Graph Combinatorial Optimization (importance: 8, dev: 5)
Xetrieval: Explaining Dense Retrieval through Feature Elicitation (importance: 15, dev: 20)
MLEvolve: A Self-Evolving Framework for Automated Machine Learning Algorithm Discovery (importance: 20, dev: 30)
ManimAgent: Self-Evolving Multimodal Agents for Visual Education (importance: 15, dev: 30)
How Far Can Root Cause Analysis Go on Real-World Telemetry Data? (importance: 20, dev: 40)
DGA$_2$D: Directed Graph-Guided Automated Algorithm Design with Large Language Models (importance: 10, dev: 10)
Evaluating Rational Contracting in Natural Language (importance: 10, dev: 5)
From Faulty Memories to Corrected Actions: Dependency-Guided Rollback Repair for Memory-Augmented Agents (importance: 25, dev: 40)
When Self-Consistency Backfires: Majority Vote Hurts the Majority of Hard Science Problems for Small LLMs (importance: 15, dev: 10)
D2R2: Discrete Diffusion with Regulation Reinforcement for Single-Cell Perturbation Prediction (importance: 5, dev: 5)
TRCA: Transition-wise Rubric Credit Assignment for Long-horizon LLM Agents (importance: 20, dev: 20)
When Prediction Error Is Not Enough: Evaluating Nuisance-Function Prediction for Causal Estimation (importance: 10, dev: 15)
Interface-Induced Trajectory Censoring (importance: 30, dev: 55)
LEBGen: An LLM-Enhanced Bayesian Network Framework for Few-Shot Travel Survey Data Generation (importance: 5, dev: 5)
Kernel-Managed Shared Memory for System-Wide Personalization (importance: 25, dev: 40)
ChatT2: An Adaptive Framework for Developing a Large Language Model-Based Agent for Natural Product Domain Research (importance: 5, dev: 15)
Driving Epidemic Models with AI Agents: the Epydemix Agent Framework (importance: 10, dev: 35)
Compositional Safety Failures in Harness Evolution: Identification and Runtime Monitoring (importance: 35, dev: 45)
R$^2$ Flow: Recursive Self-Improvement via Recursive Skill Evolution (importance: 25, dev: 35)
EvoSteer: Online Self-Evolving Graph Orchestration via Reference-Anchored Credit Assignment (importance: 25, dev: 35)
JusticeAxis: Benchmarking Legal Judgment between Rigid Rule Application and Ungrounded Discretion (importance: 10, dev: 5)
MemMux: Runtime Verification and Honest Resource Attribution for Fleets of Parallel Coding Agents (importance: 40, dev: 75)
FreeEvolve: Learning to Evolve Beyond Fixed Loops (importance: 25, dev: 35)
RetryBench: Separating Retry Permission from Recovery Outcomes in GPT Tool Use (importance: 40, dev: 65)
CLIVO: Differentiable Cluster-Based Instruction Mixing with Learned Validation Objectives (importance: 10, dev: 5)
Realtime-Bench: Can Frontier Models Solve Real-time Interactive Problems? (importance: 15, dev: 10)
Improving Environment-Agnostic Agents: The Role of Scaffolding (importance: 35, dev: 50)
R$^2$Orch: Recursive Self-Improvement for Agent Orchestration via Self-Play within a Skill-Internalization Loop (importance: 20, dev: 30)
TorchTPU - A Hybrid Eager/Compiled Architecture for Native PyTorch Execution on TPU (importance: 20, dev: 70)
Pluralistic Alignment in Large Language Models: Objectives, Human References, and Preference Representations (importance: 12, dev: 5)
When Should AI Chime In? Benchmarking Timely Contribution of Proactive Agents in Multi-Party Chat (importance: 15, dev: 10)
REFLEX: RNA Ensemble Generation via Flexibility-Calibrated Stochastic Bridge (importance: 5, dev: 3)
WhatIf-Bench: Benchmarking Intervention-Aware Time Series Forecasting (importance: 10, dev: 10)
RSI Control: Can AI Models Perform Covert Sabotage in the Self-Improvement Process? (importance: 35, dev: 15)
CrystalAR: Autoregressive Modeling for Symmetric Crystal Generation (importance: 5, dev: 3)
From Comparison to Selection: Enhancing Cross-Scale Generalization of Pointer Networks Trained with Reinforcement Learning for for Capacitated Vehicle Routing Problem (importance: 8, dev: 5)
Dancing Diffusion: Learning Continuous Programs for Image Generation and Editing (importance: 10, dev: 5)
StochIKM: Automated High-Fidelity Reconstruction of Individual Patient Data from Published Kaplan-Meier Curves (importance: 5, dev: 10)
AgentMonopoly: Behavioral Patterns and Persona Control in LLM Agents (importance: 12, dev: 10)
TraceForge: Distilling with Your Own Thinking (importance: 30, dev: 30)
Evolutionary Deep Nets for Non-Intrusive Load Monitoring (importance: 5, dev: 5)
Hearing the shape of an arena with spectral swarm robotics (importance: 5, dev: 5)
ACTIVA: Amortized Causal Effect Estimation via Variational Autoencoders (importance: 10, dev: 5)
The Mathematical Theory of Double Descent as Shrinkage, not Overfitting: Matching Bounds for Path-Regularized Multilayer Neural Networks (importance: 8, dev: 3)
Does More Compute Make Every Adversarial Attack Transferable? (importance: 15, dev: 10)
How to Backdoor Image Knowledge Distillation (importance: 12, dev: 10)
DetoxAI: a Python Toolkit for Debiasing Deep Learning Models in Computer Vision (importance: 30, dev: 50)
MAVOS-DD: Multilingual Audio-Video Open-Set Deepfake Detection Benchmark (importance: 25, dev: 15)
ShIOEnv: A Command Evaluation Environment for Grammar-Constrained Synthesis and Execution Behavior Modeling (importance: 35, dev: 45)
MoVoC: Morphology-Aware Subword Construction for Geez Script Languages Corrected version (importance: 20, dev: 15)
Interpreting vision transformers via residual replacement model (importance: 40, dev: 25)
How Well Does Mathematical Reasoning Transfer Across Languages in English-Centric LLMs? (importance: 35, dev: 15)
FlowCritic: Bridging Value Estimation with Flow Matching in Reinforcement Learning (importance: 25, dev: 10)
Revisiting GAN with Bayes-Optimal Discrimination (importance: 20, dev: 5)
Deep Learning-Enhanced Real-Time Wi-Fi Sensing Through Single Transceiver Pair (importance: 15, dev: 5)
FLAWS: A Benchmark for Error Identification and Localization in Scientific Papers (importance: 40, dev: 20)
MarkTune: Improving the Quality-Detectability Trade-off in Model-Embedded LLM Watermarking (importance: 35, dev: 20)
LLM-Auction: Generative Auction towards LLM-Native Advertising (importance: 25, dev: 10)
Calibrated uncertainty quantification for prosumer flexibility aggregation in ancillary service markets (importance: 10, dev: 0)
SpatialMem: Metric-Aligned Long-Horizon Video Memory for Language Grounding and QA (importance: 30, dev: 10)
Exact and Asymptotically Complete Robust Verifications of Neural Networks via Ising Solvers (importance: 25, dev: 10)
Near-Optimal Regret for KL-Regularized Multi-Armed Bandits (importance: 20, dev: 5)
Place-it-R1: Unlocking Environment-aware Reasoning Potential of MLLM for Video Object Insertion (importance: 20, dev: 5)
Optimistic Policy Regularization: Elite-Statistic Reward Shaping (importance: 20, dev: 10)
Markovian Generation Chains in Large Language Models (importance: 30, dev: 10)
Brittlebench: Quantifying LLM robustness via prompt sensitivity (importance: 35, dev: 30)
A Sociolinguistic Analysis of Automatic Speech Recognition Bias in Newcastle English (importance: 15, dev: 5)
Measuring the Permission Gate: A Stress-Test Evaluation of Claude Code's Auto Mode (importance: 50, dev: 80)
AtManRL: Towards Faithful Reasoning via Differentiable Attention Saliency (importance: 25, dev: 10)
Graph Memory Transformer (GMT) (importance: 30, dev: 10)
Unifying Goal-Conditioned RL and Unsupervised Skill Learning via Control-Maximization (importance: 20, dev: 5)
Benchmarking Compositional Generalisation for Machine Learning Interatomic Potentials (importance: 15, dev: 0)
Disentangled Sparse Representations for Concept-Separated Diffusion Unlearning (importance: 25, dev: 5)
Manifold Sampling via Entropy Maximization (importance: 15, dev: 5)
Learning Unbiased Permutations via Flow Matching (importance: 15, dev: 5)
Stop Drawing Scientific Claims from LLM Social Simulations Without Robustness Audits (importance: 25, dev: 5)
FormalASR: End-to-End Spoken Chinese to Formal Text (importance: 25, dev: 15)
When Should LLMs Reason? Adaptive Inference Routing via Entropy Dynamics (importance: 35, dev: 35)
How Far Will They Go? Red-Teaming Online Influence with Large Language Models (importance: 25, dev: 5)
Confidence and Calibration of Activation Oracles for Reliable Interpretation of Language Model Internals (importance: 25, dev: 10)
Semantic Optimal Transport for Sparse Autoencoder Feature Matching and Circuit Compression (importance: 20, dev: 5)
Rethinking Molecular Text Representations for LLMs: An Empirical Study (importance: 20, dev: 5)
Behavioral Consistency in Machine Learning (importance: 40, dev: 45)
Hacking Generative Perplexity: Why Unconditional Text Evaluation Needs New Metrics (importance: 20, dev: 5)
Measuring Pretraining Contamination in Public Medical Vision-Language Benchmarks (importance: 20, dev: 5)
Beyond Prompt and Parser Engineering: Soft-Prompt Tuning for Automated LLM Evaluation (importance: 30, dev: 25)
Perceive, Interact, Reason: Building Tool-Augmented Visual Agents for Spatial Reasoning (importance: 25, dev: 15)
SDS-LoRA: Overcoming Anisotropic Gradient Scaling in Low-Rank Adaptation (importance: 30, dev: 25)
Descriptor: Certus Caliber Classification Gunshot Dataset (C3GD) (importance: 10, dev: 0)
MotionHalluc: Diagnosing Kinematic Hallucinations in Fine-Grained Motion Reasoning (importance: 20, dev: 5)
Latent Bridges for Multi-Table Question Answering (importance: 25, dev: 20)
ReactiveBFM: Reactive Closed-Loop Motion Planning Towards Universal Humanoid Whole-Body Control (importance: 20, dev: 5)
PosterHarness: Turning Scientific Poster Generation into an Auditable Instruction-Following Benchmark (importance: 20, dev: 15)
Robustness Verification of an Autonomous Underwater Vehicle-based Plankton Classifier (importance: 10, dev: 0)
AirPASS: Over-the-Air Federated Learning via Pinching Antenna Systems (importance: 10, dev: 0)
Context-Aware Subtitle Translation on CPU: Deployment-Driven Training, Evaluation, and Latency Design (importance: 35, dev: 30)
WhisperRec: Latent Reasoning for Efficient Foundation Recommendation Models (importance: 25, dev: 10)
ARES: Adaptive Reasoning-Effort Steering for PPA- and Cost-Aware RTL Optimization with LLM Agents (importance: 25, dev: 20)
TORUS: A Unified Closed-Loop Benchmark for Unified Audio Models (importance: 20, dev: 10)
Principles of Robot Autonomy (importance: 10, dev: 5)
Beyond Naturalness: Probing Automated Text-To-Speech Evaluators on Linguistically Grounded Dimensions (importance: 20, dev: 5)
Rubric Dropout: A Simple Way to Mitigate Reward Hacking in Rubric-Based RL (importance: 35, dev: 20)
Mimir: A Sub-2B Hierarchical Reasoning Model Trained on Permissible Data Punches Above Its Weight Class (importance: 40, dev: 15)
Interrupting the Loop: Periodic Subject Changes Raise Judged Surprise and Connection in Base Language Models (importance: 15, dev: 0)
Moving the Mean Toward the Known Good, Not Beyond It: What Inference-Time Interventions and Weight Consolidation Buy in Open-Ended Generation (importance: 25, dev: 15)
FluxVLA Engine: A One-Stop VLA Engineering Platform for Embodied Intelligence (importance: 25, dev: 35)
Trajectory Learnability for Offline On-Policy Distillation with Imperfect Teachers (importance: 25, dev: 10)
Initial Evaluation of Potential Bias in Coverage of Humans in Wikidata (importance: 20, dev: 15)
The Price of Safety: Benign-Case Utility and Token Overhead of Memory-Poisoning Defenses in LLM Agents (importance: 35, dev: 35)
FlashBoB: I/O-Efficient Exact Backward-over-Backward for Softmax Attention (importance: 35, dev: 40)
Operator Packages, Proposer Strength, and Construction-Family Plateaus in Office-Scale Verified Search (importance: 25, dev: 15)
CollabFlow: Recursive Self-Improvement of Agent Collaboration (importance: 30, dev: 25)
Scale and Selection: What Makes Automatic Harness Evolution Work for Visual-Interface Robot Agents (importance: 35, dev: 35)
PACT: End-to-End Learning of Human Pose, Contacts, and Forces from Video (importance: 15, dev: 5)
Where Polanyi Meets Galois: Tacit Knowing as an Algebro-Deterministic Operation, Toward Re-rooting AI-Generated Science (importance: 10, dev: 10)
Beyond LLM Judge: Mechanical Verification for Agentic Autoformalization (importance: 35, dev: 25)
WorldCast: Exploring Intelligent Visual Generation in World State Prediction (importance: 15, dev: 5)
Qwen-Audio-Jev: A Unified Interface for Audio Decisions (importance: 25, dev: 25)
VEILED-SH: Vulnerability Expression through Indirect Language Evaluation Dataset (Self-Harm) (importance: 20, dev: 10)
S$^2$-MAS: A Society-Scale Multi-Agent Simulation Framework with 500 Million Household Persona Agents (importance: 20, dev: 15)
OpenMAP-HIE: Pathology Informed Score-based Diffusion Model for Unsupervised Neonatal Hypoxic-Ischemic Encephalopathy Lesion Segmentation (importance: 10, dev: 0)
Parallelizable Probabilistic Latent Reasoning for Diverse Reasoning Trajectories (importance: 30, dev: 10)
Overview of CHIP 2026 Shared Task 1: Patient-Specific Phenotype Concept Recognition in Full-Text Biomedical Literature (importance: 5, dev: 5)
Robot Generalization Beyond Volume: The Diversity Path via Zero-Shot Capabilities (importance: 20, dev: 5)
Code Comprehensor: Hierarchical Documentation and Cross-Modal Retrieval for Large Repositories (importance: 40, dev: 75)
Spectral Uncertainty Decomposition for Sparse Bayesian Learning (importance: 10, dev: 0)
AI Mindmeld: A Latent Interconnect Protocol (importance: 30, dev: 25)
How Prompt Style Affects AI Accuracy and Speed in Competitive Programming (importance: 35, dev: 45)
GANet: A Lightweight Group Attention Network for Plant Disease Identification on Embedded Systems (importance: 5, dev: 5)
Benchmarking Model Compression Techniques Across Heterogeneous Edge Platforms: A Predictive Maintenance Use Case for Rotating Machinery (importance: 20, dev: 35)
A Quadratic Lens on Muon's Polar Update: Benefits, Pitfalls, and the Role of Newton-Schulz (importance: 35, dev: 15)
Spike-Aware Gradient-Based Mixed-Precision Quantization of Spiking Neural Networks for Vision Applications (importance: 15, dev: 10)
CTRL-FM: Inference-Time Structural Control for Flow-Based Virtual Sensing under Long-Horizon Node Outages (importance: 10, dev: 5)
Bayesian Fixed-Confidence Best-Arm Identification with Abstention (importance: 10, dev: 0)
Escaping the Phylogenetic Trap: Persistent Homology Enables Cross-Species Ion Channel Classification via Multimodal Graph Neural Networks (importance: 5, dev: 5)
Memory-Augmented Online Test-Time Adaptation for EEG Motor Imagery via Perturbation Consistency and Confidence Calibration (importance: 10, dev: 0)
Towards Faithful Data Selection: Statistical Limits and Scalable Algorithms (importance: 25, dev: 15)
Stability Analysis of Signed Graphs under Structural Perturbations (importance: 5, dev: 0)
DiffLTM: Continuous Diffusions Large Tabular Models for Conditional Generation (importance: 15, dev: 15)
Optimal Multitask Representations Under a Rate Constraint: An Information Bottleneck Analysis (importance: 20, dev: 5)
NEST: Nested Embedding Representation for Multiscale Clustering (importance: 15, dev: 10)
When News Gets Old: Portfolio Allocation with Reinforcement Learning and Large Language Models (importance: 20, dev: 5)
Evaluating Type-1 Diabetes Digital Twins by Decision Quality (importance: 10, dev: 0)
RiboBridge: protein-conditioned de novo RNA binder structure design (importance: 10, dev: 0)
Learning Domain Trajectories with Flow Matching for Gradual Domain Adaptation (importance: 15, dev: 5)
An Agent Architecture with Tools and Memory for Selective Correction of Wrist PPG Heart Rate Estimates (importance: 15, dev: 15)
Solvable models of learning to pursue a moving target (importance: 15, dev: 5)
Federated Graph Retrieval-Augmented Generation (importance: 30, dev: 35)
Width and Depth Limits Commute During Training (importance: 20, dev: 10)
Random Remasking Scales Masked Diffusion Inference (importance: 25, dev: 10)
Wavenumber-Aware Deep Learning for Robust Analysis of Raman Spectra (importance: 10, dev: 5)
Visualizing and Interpreting Temporal Care Team Dynamics for Cancer Survival Risk Analysis (importance: 8, dev: 3)
Primal-Dual Representation Learning for Low-Rank Constrained MDPs (importance: 15, dev: 5)
Adaptive Deadline- and Energy-Aware Caching Algorithms in Dynamic Multi-UAV Edge Networks (importance: 5, dev: 3)
One Pretrained Decision Model Is Enough for Training-Free Generalist Graph Anomaly Detection (importance: 20, dev: 8)
Gandalf: Flash Shall Not Die! Increasing Flash Storage Lifetime for KV Cache Offloading in LLM Serving (importance: 35, dev: 45)
Trace2World: Reconstructing App Worlds from Real Interaction Traces for Scalable GUI Agent Training (importance: 30, dev: 35)
GIFT: Gradient Informed Filtering for Distillation Augmented Reasoning RL (importance: 20, dev: 10)
A Computationally Efficient and Rate-Optimal Algorithm for Adversarial Linear MDPs with (Aggregate) Bandit Feedback (importance: 12, dev: 3)
Is spectral predictability enough? Testing {\Omega}-matching for synthetic pretraining under distribution shift (importance: 12, dev: 3)
Lean Finder V2: Semantic Search for Mathlib That Disambiguates User Queries (importance: 30, dev: 40)
Reverse-KL Policy Distillation: Statistical Guarantees and Limits (importance: 15, dev: 5)
Training Mixed-Monotone Neural Networks for Verification Without Set Propagation (importance: 12, dev: 5)
Joules-to-GDP: How Much Economic Value Do AI Systems Produce per Unit of Energy? (importance: 20, dev: 10)
One Teacher, Many Views: Reducing Reference Sensitivity in On-Policy Self-Distillation (importance: 18, dev: 5)
Inherited, Not Invented: Tracing the Provenance of Wait in Reasoning Models (importance: 25, dev: 10)
Tracing Subliminal Learning Through Pretraining (importance: 25, dev: 5)
Building Presentation Intelligence for High-Quality Slide Generation (importance: 15, dev: 10)
Same molecule, Different Knowledge: Cross-Identifier Binding Failures in LLMs For chemistry (importance: 10, dev: 5)
Random Initialization as a Baseline for Time Series Foundation Models (importance: 15, dev: 5)
Ground-Truth-First Benchmarking of Reinforcement Learning via Converse Optimality: From Smooth, Fully Observed Systems to Non-Smooth or Partially Observed Tasks (importance: 12, dev: 8)
Doubly Robust Distributionally Robust Offline Contextual Pricing (importance: 8, dev: 3)
Energy and Latency First: How Deployment Constraints Reverse Time-Series Model Selection (importance: 15, dev: 10)
EEG-Captioner: Grounding EEG-to-Language Generation in Fundamental Signal Properties (importance: 10, dev: 3)
XALIGN: A Human-centric Visual Analytics Approach to Understanding AI Explanations (importance: 15, dev: 10)
Federated Differentially Private Optimization via Low-Rank Preconditioning (importance: 12, dev: 5)
When Forgotten Facts Return: Relearning-Aware Unlearning of Knowledge Graph Embeddings via Tangent Kernels (importance: 12, dev: 5)
nnInteractive Agent: General-Purpose AI Agents as 3D Image Annotators (importance: 30, dev: 25)
Transformers with CoT Discover Heavy-Ball Acceleration Through Training (importance: 15, dev: 3)
What Comes Next and Why: Interpretable Next-Event Prediction with Neuro-Symbolic Rules (importance: 12, dev: 5)
Is High Local Accuracy Enough? Geometric Consensus with Support-Adaptive Statistical Personalization for Federated Learning (importance: 10, dev: 3)
Towards Field-Deployable Wearable IMU Systems for Occupational Activity Recognition (importance: 8, dev: 2)
Inside-JEPA: From Prediction to Relational Grounding in Latent World Models (importance: 22, dev: 3)
Evolving Federated Density Peak Clustering with Shuffler-Based Anonymization (importance: 8, dev: 3)
Local Discrimination of Binary Classifiers via Localized AUC Functional (importance: 8, dev: 3)
QuaNAS: Uncertainty-Aware Reuse of Earlier Searches for Query-Efficient Neural Architecture Search (importance: 15, dev: 10)
Deep ReLU Networks Can Fail the Best Approximation Property for Continuous Target Functions (importance: 12, dev: 2)
DISC: Domain-Integrated Similarity Cost for Time-Series Dataset (importance: 12, dev: 8)
The Bayesian Learning Rule Beyond KL Geometry (importance: 10, dev: 2)
Extreme Compression of Per-Layer Embeddings via Cross-Layer Latent Vector Quantization (importance: 35, dev: 25)
Identifiability from Stationary Covariance in Linear SDEs with Reciprocal Interactions and Latent Variables (importance: 6, dev: 2)
The Value of System 1 Models in Medicine (importance: 10, dev: 3)
Harmful Responses Are Jailbreak Templates: Transferring Autoregressive Jailbreaks to Diffusion Language Models (importance: 30, dev: 15)
SQuAT: Spherical Lattice Vector Quantization-Aware Training (importance: 25, dev: 10)
Drift Says When, Budget Says How Much: Rate-Controlled KV Re-selection for Long-Form Multimodal Generation (importance: 22, dev: 15)
Multi-Token Prediction and Alternative Tokenization Improve Learning Latent Multi-Hop Graph Search at Scale (importance: 15, dev: 5)
RheOPT: Few-Shot Non-Newtonian Flow Prediction via Newtonian Pretraining of RheOFormer (importance: 5, dev: 2)
Evaluating the Applicability of LLM Inference Optimizations on Apple Silicon (importance: 40, dev: 60)
Piano2Pop: Expanding Piano Demos into Pop Productions Without Aligned Supervision (importance: 12, dev: 3)
PYTHIA, a transferable machine learning framework for simultaneous defect and adsorption site optimization (importance: 6, dev: 2)
Same Luck, Different Choice: Twin-World Credit Assignment for Agentic Reinforcement Learning (importance: 25, dev: 15)
Shift-aware score transport for privacy auditing without retraining (importance: 12, dev: 5)
Fine-Grained Analysis of Catastrophic Forgetting in Continual Learning via Algorithmic Stability (importance: 12, dev: 3)
Diffusion Language Models for Perplexity-Aware Black-Box Attacks (importance: 25, dev: 8)
On the Token Collapse Phenomenon in Textual Graph Soft Prompting (importance: 12, dev: 5)
Rethinking Sequence Compression with Test-time Regression (importance: 20, dev: 8)
Action Learning for Scientific Simulations (importance: 8, dev: 2)
CRANE: Post-Hoc Counterfactual Retrieval for Multimodal Classifiers (importance: 10, dev: 3)
Efficient and Optimal Learning of Discrete Distributions with Person-Level Privacy (importance: 6, dev: 2)
OP-Print: Tracing Membership Imprints in On-Policy Self-Distillation (importance: 15, dev: 5)
BeNice! Closing Looping's Inference-Memory Gap via KV-Sharing (importance: 18, dev: 10)
A Fix for Softmax and Cross-Entropy (importance: 35, dev: 55)
Do Machine Learning-Guided Proactive Full Capacity Protocols Mitigate Emergency Department Overcrowding? A Simulation-Based Evaluation (importance: 5, dev: 2)
baseVultra: Compact Nonlinear Recurrence with Reallocated Capacity (importance: 10, dev: 5)
SIGNN++: Robust Graph Neural Star Identification on Earth (importance: 5, dev: 2)
Differentially Private Synthetic Data via Generalized Moment Matching (importance: 10, dev: 3)
EgoHand: Monocular 3D Hand Reconstruction from Egocentric Video (importance: 15, dev: 5)
SciFusion: Physics-Guided Compositional Diffusion for Multimodal 3D Scientific Image Generation (importance: 8, dev: 3)
AQL-Lite: Enabling Efficient Knowledge Acquisition for Arabic LLMs via Data-Centric Continual Pretraining (importance: 10, dev: 3)
On the Adversarial Vulnerability of Neural Receivers with Unrolled LDPC Decoders (importance: 8, dev: 3)
A Transformer-based Approach for Real-time Multi-agent Path Planning with Varying Initial/Final Conditions (importance: 12, dev: 5)
Flow-based Spectral Kernel Learning for Nonstationary Attention (importance: 12, dev: 5)
Learning to Incentivise: Using Reinforcement Learning for Sustainable Urban Mobility (importance: 8, dev: 3)
FIGFLEX: A Mixed-Precision FP-INT GEMM Accelerator Co-Designed with Low-Bit LLM Quantization (importance: 15, dev: 10)
Fast Iterative Combinatorial Auctions via Deep Reinforcement Learning (importance: 8, dev: 3)
AIRO: A Streaming-Based Machine Learning Framework for Threat Detection and Security in Industrial Internet of Things (importance: 8, dev: 5)
Amplifying Latency and Cost of LLM Inference via Rowhammer Attack (importance: 25, dev: 15)
xVGAE: A Hierarchical Variational Graph Autoencoder for Exchangeable Graphs (importance: 8, dev: 2)
Continuous-depth Deep Gaussian Processes (importance: 6, dev: 2)
Equilibria and Learning in Competitive Market Making (importance: 5, dev: 1)
Bridging Number-Theoretic Primality Testing And Algorithmic Optimization: An Adaptive Cryptographic Framework (importance: 5, dev: 5)
PAIR: Preference-Aligned Intent Retrieval for Premium People Recommendation at LinkedIn (importance: 15, dev: 8)
Random Projections for Spectral Algorithms: Optimal Rates in Sobolev Norms (importance: 8, dev: 2)
ReDence: A Hybrid Agentic and Decision-Model Framework for Rapid, Evidence-Grounded Detection of Modern Social Engineering Scams (importance: 12, dev: 8)
Joint or Per-Option? Compact Encoders for Closed-Set Decisions from 17 Million to 1 Billion Parameters (importance: 25, dev: 20)
Diagnostic Loss: Asymmetric Supervision for Radiology Report Generation (importance: 8, dev: 2)
NN-Based Transistor Modeling for Ultra-Fast SPICE Simulation and Standard Cell Library Characterization (importance: 12, dev: 8)
DRAFT: Decoding and Reconstructing 3D Room Layouts from fMRI and Camera Trajectories (importance: 5, dev: 1)
Delta N-gram: Forward-Only Continual Writing into a Language Model's Native N-gram Memory (importance: 18, dev: 10)
GM-GReFEL: Geometry-Guided Spatiotemporal State-Space Learning for Robust Dynamic Facial Expression Recognition (importance: 8, dev: 2)
Prediction Uncertainty Controlled Computer Model Calibration (importance: 8, dev: 3)
Memory before Goal: Role of Pre-History Memory for Humanoid Loco-Manipulation (importance: 12, dev: 4)
Track finding and particle identification performance with particle physics foundation models (importance: 6, dev: 2)
RaSFNet: Ray-aligned and Scan-aware Fusion Network for LiDAR Semantic Segmentation (importance: 6, dev: 2)
Muting Tokens: Exploring Token Sparsification for Efficient Audio Transformers (importance: 10, dev: 4)
Learning Robot Behavior Priors from Non-Robotic Synthetic Data (importance: 12, dev: 3)
Data-Driven Vision-Language Models for Remote Sensing: A Survey (importance: 25, dev: 5)
Deep Learning for Random Objects in Metric Spaces: Beyond Manifold Assumptions (importance: 15, dev: 3)
Metric Conformal Prediction via the Conditional Expected Local Radius (importance: 12, dev: 2)
Graph-Recurrent Learning of Emergent Collective Dynamics (importance: 10, dev: 2)
AfterDeleteBench: Auditing Post-Deletion Information Persistence in LLM Agents (importance: 35, dev: 40)
Convex Equilibrium Potentials for Monge--Amp\`ere: Curvature Bounds and A Posteriori Evaluation (importance: 8, dev: 2)
Dimension-free regimes for unadjusted Langevin in ridge-separable models: A subspace robust Wasserstein analysis (importance: 10, dev: 2)
Structure-preserving reduced-order models of thin panels in supersonic and hypersonic flow (importance: 8, dev: 2)
Nonparametric autoregressive estimation of probability distributions with parallel gradient boosting (importance: 15, dev: 8)
Direct Inference on Admissible Sets (Diona) (importance: 15, dev: 8)
LACToS: Local Anomaly Detection via Conditional Typicality of Superpixels (importance: 15, dev: 5)
Beyond Best-of-$k$: Prior-Consistent DAPS for Ambiguous Nonlinear Inverse Problems (importance: 12, dev: 3)
SpectralCanary: Embarrassingly Simple Low-Frequency Traces for T2I Dataset-Use Auditing (importance: 18, dev: 5)
Reliable AI-Assisted Decisions via Selective Auditing (importance: 20, dev: 15)
Phase Transition from Random Features to Fully Trainable Networks (importance: 10, dev: 3)
Sensorimotor Alignment: A Platonic Proxy for Embodied Policy Success (importance: 12, dev: 3)
On Sex, Evolution, and the Multiplicative Weights Update Algorithm (importance: 5, dev: 1)
Are EEG Sequences Time Series? EEG Classification with Time Series Models and Joint Subject Training (importance: 8, dev: 3)
The Fenchel-Young Generalization of Bayesian Principles (importance: 10, dev: 2)
Fragility-aware Classification for Risk Characterization and Robust Learning (importance: 12, dev: 4)
Multi-View Causal Discovery without Non-Gaussianity: Identifiability and Algorithms (importance: 10, dev: 2)
Analysis of On-policy Policy Gradient Methods under the Distribution Mismatch (importance: 12, dev: 3)
Aggregation of Published Non-Uniform Axial Power Data for Phase II of the OECD/NEA AI/ML Critical Heat Flux Benchmark (importance: 5, dev: 2)
Sparse Equation Matching: A Derivative-Free Learning for General-Order Dynamical Systems (importance: 8, dev: 2)
Energy Estimation for Virtual Servers Using Resource Utilization and Machine Learning (importance: 15, dev: 30)
Robust Tangent Space Estimation via Laplacian Eigenvector Gradient Orthogonalization (importance: 10, dev: 3)
To Distill or Decide? Understanding the Algorithmic Trade-off in Partially Observable Reinforcement Learning (importance: 12, dev: 4)
Specter: Fast MoE Inference via Quantization-Enabled Multi-Step Speculative Offloading (importance: 25, dev: 25)
How Does RL Post-training Induce Skill Composition? A Case Study on Countdown (importance: 15, dev: 5)
Verification of Unknown Dynamical Systems via Autoencoder Latent Space (importance: 8, dev: 4)
TabClustPFN: A Prior-data Fitted Network for Tabular Data Clustering (importance: 12, dev: 8)
Are Two LLMs Better Than One? A Student-Teacher Dual-Head LLMs Architecture for Pharmaceutical Content Optimization (importance: 12, dev: 10)
K-Means as the Zero-Temperature Limit of Normalized RBF Assignments: Rates, Trajectories, and Exact Recovery with Entmax (importance: 10, dev: 3)
An Optimal Control Approach To Transformer Training (importance: 12, dev: 5)
Show Me What You Don't Know: Efficient Sampling from Invariant Sets for Model Validation (importance: 15, dev: 10)
Unrestrained Simplex Denoising for Discrete Data. A Non-Markovian Approach Applied to Graph Generation (importance: 10, dev: 3)
Super Agents and Confounders: Influence of surrounding agents on vehicle trajectory prediction (importance: 10, dev: 3)
One Algorithm, Two Goals: Dual Scoring for Parameter and Data Selection in LLM Fine-Tuning (importance: 25, dev: 30)
Efficient Adjoint Matching for Fine-tuning Diffusion Models (importance: 12, dev: 8)
IGADA-IoT: IoT Sensor Energy Optimization in Wireless Sensor Networks Driven by Automatic Data Augmentation (importance: 5, dev: 5)
Can Entry-Wise Clipping Give Spectral Control of Stochastic Gradients? (importance: 25, dev: 25)
SoundnessBench: Can Your AI Scientist Really Tell Good Research Ideas from Bad Ones? (importance: 20, dev: 10)
Measuring, Localizing, and Reducing the AI-Like Writing Style of Post-Trained LLMs (importance: 20, dev: 8)
LSVD: Loss-Aware Low-Rank Approximation for Efficient Low-Precision Vision-Language Models (importance: 18, dev: 20)
Measuring Model Robustness via Fisher Information: Spectral Bounds, Theoretical Guarantees, and Practical Algorithms (importance: 12, dev: 5)
How Far Can Sharpness and Complexity Jointly Explain Generalization? (importance: 10, dev: 3)
LLM-Guided Transportation Hub Capacity Planning with Textual Business Inputs (importance: 8, dev: 5)
Learning Discrete Riemannian Metrics for Physical Fields with Cochain-Frame Equivariance (importance: 8, dev: 3)
Do Tabular Foundation Models Know Physics? Contamination, Units, and the Deterministic Limit (importance: 12, dev: 8)
Multi-center Medical Data Mining with FL-Net - A One-stop Shop for Federated Learning (importance: 12, dev: 15)
Extending FunctionGemma for Practical On-Device Mobile Function Calling (importance: 20, dev: 25)
Residual Correlation as a Diagnostic for Joint-Uncertainty Gains from GP Coregionalisation (importance: 6, dev: 2)
PROACT-Agent: Progressive Runtime Oversight and Active Circuit-breaking for Real-Time Safety (importance: 25, dev: 25)
UnlearningSoup: Is Repeated Tuning Necessary for Large Language Model Unlearning? (importance: 15, dev: 8)
DashVMC: Real-Time Discrete World Model Control in Geometry Dash (importance: 8, dev: 3)
Redundancy Meets Synergy: Dependency-aware Expert Selection for MoE via Submodular Optimization (importance: 15, dev: 12)
Adversarial Reinforcement Learning for Robust Planetary Landing (importance: 8, dev: 3)
Predictable Scaling: Efficient Model Exploration via Scaling-Law Prediction (importance: 20, dev: 12)
Generative Retrieval for Reverse Target Fishing with a Proteome-Scale Benchmark (importance: 8, dev: 3)
NADS: Navigator-Guided Data Selection for Mitigating Catastrophic Forgetting in Fine-Tuning (importance: 18, dev: 15)
Poisoning-Based Backdoors Are Not Cryptographic (importance: 15, dev: 5)
Present but Hard to Read: Probing Latent Reasoning with Minimal Pairs (importance: 12, dev: 4)
Optimal Clustering of Multi-layer Networks under Mixture Block Models (importance: 5, dev: 1)
Reducing QPU Calls in Variational Quantum Algorithms via Classical Offloading of Gradient Computation (importance: 6, dev: 3)
Eluder-Dimension-Free Regret for Stage-Wise Constrained Contextual Bandits via Offline Regression (importance: 5, dev: 1)
Diffusion-Based Hypothesis Testing and Change-Point Detection (importance: 5, dev: 2)
Few-shot Classification as Multi-instance Verification: Effective Backbone-agnostic Transfer across Domains (importance: 12, dev: 8)
On the Gradient Domination of the LQG Problem (importance: 6, dev: 2)
Meta-learning of Gibbs states for many-body Hamiltonians with applications to Quantum Boltzmann Machines (importance: 6, dev: 1)
Automatic Fine-grained Segmentation-assisted Report Generation (importance: 10, dev: 4)
Stochastic Matching Bandits with Rare Optimization Updates (importance: 5, dev: 1)
Bayesian Ambiguity Contraction-based Adaptive Robust Markov Decision Processes for Adversarial Surveillance Missions (importance: 6, dev: 2)
Penalized Fair Regression for Multiple Groups in Chronic Kidney Disease (importance: 6, dev: 2)
High-dimensional learning dynamics of multi-pass Stochastic Gradient Descent in multi-index models (importance: 6, dev: 2)
Learning to Drive in New Cities Without Human Demonstrations (importance: 12, dev: 4)
Structured Unitary Tensor Network Representations for Circuit-Efficient Quantum Data Encoding (importance: 5, dev: 2)
CoarseSoundNet: Building a reliable model for ecological soundscape analysis (importance: 5, dev: 2)
Evaluating Neural Probabilistic Forecasts of Sparse Earthquake Counts: What Survives a Careful Evaluation (importance: 6, dev: 2)
SAHG: Sector-Anisotropic Hyperbolic Graph Model for Social Bot Detection (importance: 10, dev: 3)
Giskard : Byzantine Robust and Private Aggregation for Large-Scale Decentralized Learning (importance: 8, dev: 4)
Pathways of Visual Information Flow in Vision-Language Models (importance: 12, dev: 4)
Parallel gradient boosting for flexible estimation of conditional distributions (importance: 6, dev: 3)
What Does It Take For A Generalist Robot Policy To Adapt Using Minimal Data? (importance: 12, dev: 4)
POI Recommendation with LLM-Augmented Multi-Graph Learning and Contrastive Alignment (importance: 6, dev: 3)
Decoupling candidate dual AGN from chance superpositions in the GOTHIC survey via a deep-learning framework (importance: 3, dev: 1)
Advances in Machine Learning for Directed Evolution: A Five-Year Retrospective (importance: 6, dev: 1)
Fixed-Dimensional Latent Flow for Generating Variable-Size 3D Molecules (importance: 6, dev: 2)
Semantics Delivery Network: Rethinking Web Retrieval Infrastructure for LLM Agents (importance: 22, dev: 35)
Reinforcement Learning and Operational Research: A Review of Methodological Integration and Research Directions (importance: 8, dev: 4)
Is the Judge Its Twin? Post-training makes a model's writing more predictable but barely moves its taste, as a judge, toward predictable writing (importance: 15, dev: 10)
ATTUNE: Agentic Throughput Tuning for LLM Training (importance: 20, dev: 25)
Deficit-Driven Stratified Acquisition for Building STRATA-S12: An Open SAR-Optical Corpus for Self-Supervised Learning (importance: 4, dev: 2)
An Adaptive-Horizon Graph Neural Operator for coupled surrogate morphodynamic modelling (importance: 4, dev: 2)
ShlokaDhara: A Multi-Speaker Sanskrit Speech Corpus for Learning Meter-Aware Prosody in Text-to-Recitation (importance: 8, dev: 4)
Disclosure Begets Disclosure: A Self-Reinforcing Privacy Loop in LLM Agents with Persistent Memory (importance: 20, dev: 20)
Closed-Loop Dual-Reference Guidance for Accent Control in Cross-Lingual Voice Cloning (importance: 8, dev: 4)
RangeNav: Task-Aligned LiDAR Range-Image Pretraining for Robot Navigation (importance: 8, dev: 4)
Tractable Inference with Kolmogorov-Arnold Networks (importance: 10, dev: 4)
A Consolidated Deployment Edition of a Microservice Core-Banking Platform: Composing Services by Reference into Three Hosts (importance: 2, dev: 55)
JANUS: Agentic Cross-Language Fuzzing Harness Generation for Android Native Libraries (importance: 15, dev: 35)
Agentic Architecture of Information Systems. The AAM/5C Model: A Reference Model for Distributed Autonomy (importance: 15, dev: 30)
The Customer Is the Builder: How AI Coding Agents Are Rewriting the Economics of Software (importance: 30, dev: 40)
TrustedCode: Evaluating Large Language Models for Developing Trusted Code through Automated Program Verification (importance: 35, dev: 55)
Stale Instruction Files Override the Code in Coding Agents: A One-Sentence Fix (importance: 45, dev: 80)
From Tokens to Credits to Outcomes: Pricing AI Across Models, Applications, and Cloud Platforms (importance: 30, dev: 40)
From Natural Language to Verified C: Verification-Guided Generation with ACSL Specifications (importance: 35, dev: 60)
What Ethereum Developers Ask For, and Whether They Get It: Ten Years of Ethereum Stack Exchange (importance: 15, dev: 45)
ATLAS: A Multi-Agent Framework for Lifting Binary to Canonical Compiler IR (importance: 30, dev: 40)
When Code Gets Cheap, Evaluation Gets Scarce: An Experience Report on AI-Assisted Agile Software Engineering (importance: 40, dev: 80)
REGATTA: A Cross-Domain Measurement Framework and Fault-Injection Harness for Enterprise AI Agents (importance: 25, dev: 45)
Tracing Hallucinations in LLM Reasoning for UML Class Diagram Generation: An Empirical Taxonomy and Mitigation Study (importance: 20, dev: 30)
Typed Template Fuzzing (importance: 25, dev: 60)
Procrustes-Bering-Sheaft: Automated Model Discovery and Virtual Crash Tests for Microservices (importance: 20, dev: 60)
CAMGen: A Core Change-Aware Multi-View Framework for Commit Message Generation (importance: 25, dev: 60)
Deep Search in Real-World Web Environments (importance: 30, dev: 30)
When EVM Defaults Break: Three SDK Patterns for UTXO-Family Wallets (importance: 25, dev: 70)
ConQEC: Concolic Testing for Quantum Error-Correcting Code Implementations (importance: 15, dev: 40)
DynamiQ: Unlocking the Potential of Dynamic Task Allocation in Parallel Fuzzing (importance: 15, dev: 45)
Model of Things (MoT): A Model-Driven Approach to Streamline Cloud-of-Things Application Development (importance: 10, dev: 30)
LIDL: LLM Integration Defect Localization via Knowledge Graph-Enhanced Multi-Agent Analysis (importance: 25, dev: 65)
APO-DD: Dialogue Disentanglement via Reflective Optimization of Multi-Component Prompts (importance: 15, dev: 25)
A Multi-Dimensional, Per-Pass Empirical Study of the LLVM Optimization Pipeline (importance: 20, dev: 55)
No Edges, No Verdict: A Large-Scale Empirical Study of Declared Dependency Graphs in 78K SBOMs in the Wild (importance: 20, dev: 60)
AuditCoder: Process-Auditable Code Generation through Responsibility-Preserving Task Graphs (importance: 25, dev: 60)
The glorious @xkcdRandom Zig builtin (importance: 2, dev: 55)
Software is slow because of bad design, not slow languages (importance: 2, dev: 50)
UUIDv7 doesn't have the properties you think it does (importance: 2, dev: 65)
Please don't mess with links (importance: 2, dev: 40)
How close can pure FP get to C? (importance: 2, dev: 50)
Debunking "Faster than Rust": Malicious Benchmarks in Tov (importance: 2, dev: 45)
Emacs finally folds to my liking: kirigami, tree-sitter, hide-show, Org, XML (importance: 1, dev: 40)
Quad9 DNS Resolver to exit France (importance: 1, dev: 15)
a hacky fucked up way to portably access errno from outside of c (importance: 2, dev: 50)
What are you doing this weekend? (importance: 0, dev: 0)
a neat trick i found for reusing physical memory (importance: 2, dev: 45)
We have Traits at Home (importance: 2, dev: 45)
A whirlwind tour of systemd-nspawn containers (2025) (importance: 2, dev: 40)
examining the macOS gcc binary (importance: 2, dev: 35)
The Many Many Missed Opportunities of Sonic 06 (importance: 0, dev: 0)
The Hibernate Origin Story (Outsider Edition) (importance: 2, dev: 40)
How to Make Things, and Why (importance: 3, dev: 30)
Release OpenSSL 4.1.0 · openssl/openssl (importance: 5, dev: 75)
No Source Code? No Problem! What if you have to port a program, but all you have is a binary? (2003) (importance: 1, dev: 40)
Expose only share links of a private service with caddy (importance: 1, dev: 45)
Next-Concept Modeling: Fast, Parallel Coarse-to-Fine Text Generation (importance: 25, dev: 15)
Objects, attributes, maps, oh my (importance: 1, dev: 55)
Tell your network to explain itself and then some (importance: 1, dev: 30)
Agent Skills Explained: How to Write a SKILL.md for Claude Code, Cursor, Codex and Copilot (importance: 0, dev: 80) (いいね相当スコア: 0)
LazyFlow AI: Turn Messy Notes into Actionable Tasks with Local AI (importance: 0, dev: 30) (いいね相当スコア: 0)
Eight eMudhra Intermediate CA Keys Rested on Eight Bits (importance: 0, dev: 55) (いいね相当スコア: 0)
Tool-use agents that actually finish the job (importance: 0, dev: 60) (いいね相当スコア: 0)
The Government Says Web Dev Jobs Will Grow 5%. The Footnote Is the Interesting Part. (importance: 0, dev: 25) (いいね相当スコア: 0)
OpenAI Claude Gemini API Text Summarization Quality Cost Comparison for Game Reviews (importance: 0, dev: 40) (いいね相当スコア: 0)
Top 500 YouTube Creators in India, Bangladesh & Pakistan (2026) (importance: 0, dev: 0) (いいね相当スコア: 0)
Four x402 services, fifteen paid routes — and the cataloging mechanics nobody documents (importance: 0, dev: 60) (いいね相当スコア: 0)
Muse invite code DYGM1I: 7 questions every new user asks, answered (importance: 0, dev: 0) (いいね相当スコア: 0)
How to Debug Autonomous AI Agents in Terminal Workflows (Without Heavy Observability Stacks) (importance: 0, dev: 70) (いいね相当スコア: 0)
[Boost] (importance: 0, dev: 10) (いいね相当スコア: 0)
Plumbing AI Automation: From First Leak Call to Paid Invoice Without the Paper Chase (importance: 0, dev: 5) (いいね相当スコア: 0)
How to Use Ollama to Run AI Models on Your PC (2026) (importance: 0, dev: 35) (いいね相当スコア: 0)
Node.js Text Summarization API Cost Comparison Across Models for Media CRM (importance: 0, dev: 45) (いいね相当スコア: 0)
Text Summarization APIs: A 4-Step Portability Test for Fintech Moderation (importance: 0, dev: 45) (いいね相当スコア: 0)
How much does an AI agent cost to build and run? (importance: 0, dev: 25) (いいね相当スコア: 0)
RAG explained: how to give an LLM your own information | Xiliux Blog (importance: 0, dev: 40) (いいね相当スコア: 0)
Designing a Reliable 4-Layer LLM API for Moderation Review Chatbots (importance: 0, dev: 65) (いいね相当スコア: 0)
How do you write a coding-agent task that succeeds on the first try? (importance: 0, dev: 65) (いいね相当スコア: 0)
MiniMax Just Open-Sourced a 428B Coding Model — Can You Actually Run It? (importance: 0, dev: 60) (いいね相当スコア: 0)
what i learned building agent skills that admit what they can't check (importance: 0, dev: 60) (いいね相当スコア: 0)
MySQL as a Workflow Engine: Building an LLM Prompt Gate (importance: 0, dev: 65) (いいね相当スコア: 0)
The Model Is Not the Bottleneck. The Routing Is. (importance: 0, dev: 45) (いいね相当スコア: 0)
Claude Fable 5.1 can hide secret messages in benign images (importance: 20, dev: 15)
No coding or gamedev background, 6 weeks with Opus 5.5 + Godot — from drawings to a playable game (importance: 12, dev: 35)
Sweden’s government is blocking ChatGPT and Claude on staff computers from Monday (importance: 40, dev: 15)
a preview of AGI..😭 (importance: 0, dev: 0)
Anthropic appears to sneak in a legal loophole in Claude Code to train on "feedback" even with training off as in Claude Code's source theres a legal term that counts every "Allow edit" you press as feedback. (importance: 40, dev: 75)
This Proves That Claude Is The Best Ai. (importance: 2, dev: 10)
This is addictive. (importance: 5, dev: 20)
Another unsolved ciphers falls to AI, this one was meant to test life after death (importance: 20, dev: 10)
I couldn't understand my own jaw problems, so I had Claude Code turn my CBCT and MRI scans into a 3D x-ray film (importance: 10, dev: 20)
Choose your projects well…the gravy train will end. (importance: 10, dev: 15)
To my knowledge the first published randomised controlled trial with Claude as the intervention in patient care. 300 neurosurgical patients, NEJM AI, published this month. (importance: 35, dev: 10)
I am getting more and more tired. Is it just me? (importance: 3, dev: 10)
Claude made a documentary about life after the singularity (importance: 8, dev: 5)
WHY CLAUDE 😭 (importance: 2, dev: 5)
Claude told me to go to sleep. I thought they stopped doing this? (importance: 5, dev: 5)
Did they nerf Opus 5.5? Doesn’t look like it. LiveNerf day 29 update (importance: 30, dev: 40)
4 days and almost 4 weekly Opus limits: a multiplayer A-10 flight sim that runs in a browser tab (importance: 15, dev: 35)
I used Claude Code to build a pizzeria its own ordering app so regulars stop paying DoorDash 30%. First paying customer is live. (importance: 15, dev: 40)
Claude is Like a Senior Now (importance: 5, dev: 20)
COMPACTIONS VS HANDOFFS, let's SETTLE this right HERE! (importance: 35, dev: 65)
I'm positively surprised by the 20$ Subscription (importance: 15, dev: 35)
I got tired of checking whether Claude Code needs me. So I gave it a physical status light. (importance: 15, dev: 50)
I asked Claude if I should credit my cat in the paper I am writing... (importance: 3, dev: 5)
GitHub Actions getting really expensive. Any alternative? (importance: 5, dev: 75)
Using Codex and Claude Code every day: what's still hard once the code part works? (importance: 25, dev: 65)
Bugs I keep catching in LLM-written code, and the checks that catch them (importance: 25, dev: 75)
They replaced our two backend devs with me and Claude Code (importance: 25, dev: 45)
Your git log already knows when two agent sessions overwrote each other's work. A one-liner to ask it (importance: 30, dev: 85)
ChatGPT Plus vs Cursor Premium in 2026 — which one is actually worth it? (importance: 10, dev: 40)
Vibecoding your own inference engine is easy they said (importance: 12, dev: 40)
Models disappearing from Huggingface (importance: 15, dev: 35)
nvidia/NVIDIA-Nemotron-3.5-Bolt-E4B-4GB · Hugging Face (importance: 30, dev: 25)
Small post about what are the unlocks for RTX consumer GPUs, which ones are affected, in which tasks and which ones are feasible to do an unlock. (importance: 8, dev: 12)
everything will be local eventually (importance: 5, dev: 15)
Qwen 3.8 Flash Next Gyro Coder is out! (importance: 30, dev: 45)
Probability of backdoors in popular harnesses and LLM inference engines (importance: 20, dev: 50)
Here's where the last RTX 5090 cards are going: Massive AI conversion factory exposed in China, Founders Edition included (importance: 15, dev: 10)
MiniMax-M3.1-Flash (importance: 25, dev: 30)
Qwen3-0.6B model converted into MoE (importance: 20, dev: 35)
I ran all the popular Qwen 3.8 FN engines on a single 5090 + 128GB of RAM, here are the numbers (importance: 30, dev: 60)
[Paper] Virtual Head Attention (importance: 35, dev: 35)
Strata now supports Qwen 3.6 35B (importance: 15, dev: 40)
Does the “gpu unlock speedup” trick only work for 3090, or other 30x0 cards too? What about 40x0, 50x0? (importance: 5, dev: 10)
NanoStories Releases! (importance: 12, dev: 15)
Started to play around a bit with opencode and this happened during thinking (importance: 15, dev: 45)
Baseten releases elephant-1 (9B, MIT, Qwen3.5-9B base), a decision model that ties d1-ultra on the AA Decision Index at about a quarter of the cost (importance: 25, dev: 30)
Anyone checked gpu unlock for rtx 30-40 cards? (importance: 8, dev: 12)
is it worth to dumpsterdive? (importance: 3, dev: 5)
qwen4exp : build the QSA selection without the kq_mask by sanmai · Pull Request #30965 · ggml-org/llama.cpp (importance: 25, dev: 60)
A 4B model faster than a 395M encoder? 14 ms vs 31 ms (2.2x). Another 4B model: 150 ms (10x slower) (importance: 20, dev: 35)
agentionai/Qwen3.8-Flash-Next-Gyro-Coder-GGUF · Hugging Face (importance: 20, dev: 40)
Excited about AI agents, but also slightly terrified of letting them loose in my business! (importance: 5, dev: 15)
How much memory should an AI agent really have? (importance: 15, dev: 35)
I gave llama3.1:8b a document with hidden orders and let it try to take over my Mac. It failed every time. (importance: 20, dev: 45)
An AI agent shouldn't have more permissions than the employee it's assisting. (importance: 15, dev: 35)
A voice AI agent that sounds human but can't handle interruptions isn't production-ready. (importance: 12, dev: 25)
I connected Hermes Agent to my local Obsidian vault on a Mac mini — here's what I tested (importance: 15, dev: 30)
If your AI agent saves time but not budget, is it actually a product? (importance: 5, dev: 10)
With AI you can build almost anything, what does that mean for the future of software (importance: 5, dev: 15)
If LLMs become the universal interface, how will they discover MCP services nobody knows about? (importance: 5, dev: 20)
Meet Chronos: a database where your agent’s changes land only if they keep your rules (importance: 25, dev: 70)
Anybody figure out how to have their agents reliably find events? (importance: 5, dev: 15)
What’s the line for you between “let the agent handle it” and “a person has to touch this”? Here’s where I drew mine. (importance: 8, dev: 15)
The hooks that stopped my agent from shipping garbage (importance: 30, dev: 65)
What task did your agent get stuck on that you ended up paying a person to do? (importance: 5, dev: 20)
Looking for an Instinct invite, happy to share what we learn in return (importance: 3, dev: 5)
Anthropic's own data suggests we're becoming middle managers for AI agents (importance: 35, dev: 35)
Has anyone used the Claude Chrome extension to extract Reddit data for Meta platforms? Any risk of getting banned? (importance: 5, dev: 5)
Soroco alternatives for getting a work graph before you deploy agents (importance: 5, dev: 15)
GetJavis.EU seems legit... but is it??? (importance: 3, dev: 10)
Do you think AI will make life easier or make things worse for the next generation? (importance: 3, dev: 5)
Best app for ai viral video templates? (importance: 2, dev: 5)
What tooling do you use for a cloud-based, multi-user AI software factory? (importance: 10, dev: 45)
Best Ai I have ever used (importance: 1, dev: 0)
Altman admits the cybersecurity crisis is going to get worse: "I no longer think we're gonna be able to patch all the bugs." (importance: 40, dev: 15)
3% of critical software security bugs, in all of recorded history, have been found in the last month - "Black October" (importance: 45, dev: 25)
Do you guys remember how awful AI was like 4–5 years ago? (importance: 3, dev: 5)
National security expert: "There's a decent chance the internet ceases to exist as such, at least for a time." (importance: 30, dev: 10)
Google DeepMind founder Hassabis shortens AGI timeline from five to two to three years (importance: 35, dev: 10)
Cybersecurity people are losing their minds today. Some are calling it Black October and expect internet lockdowns soon (importance: 50, dev: 25)
Ex-OpenAI safety researcher accuses OpenAI of blatantly lying to the public (importance: 30, dev: 10)
Using OpenAI Realtime API with a Robot (importance: 15, dev: 35)
Incoming reset (importance: 10, dev: 15)
OpenAI Ads Scammed Me. Stay Away. (importance: 5, dev: 5)
Shieet, here we go again! (importance: 25, dev: 15)
You think that's air you're breathing? (importance: 2, dev: 0)
My theory on all the resets. (importance: 10, dev: 10)
OpenAI's mistake is going to cost me $1,100 (importance: 5, dev: 5)
OpenAI's Ad Pixel Tracks by Default, and European Shops Aren't Switching It Off (importance: 20, dev: 10)
We are so close to losing the ability to see what the AIs are actually thinking (importance: 30, dev: 15)
7 OpenAI updates from this week that cut your AI bill (importance: 15, dev: 55)
Your usage does not need a reset right now (importance: 5, dev: 10)
Former Chief Scientist of the UK AI Security Institute: we should stop AI progress now (importance: 30, dev: 5)
SoftBank bet on OpenAI (importance: 15, dev: 5)
I just came across Next-Concept Modeling and the idea is fascinating[R] (importance: 30, dev: 30)
AAAI 2027 phase 2 reviews discussion [D] (importance: 5, dev: 5)
Is symbolic ai dead? [D] (importance: 10, dev: 15)
How does arXiv prepublishing work with ICLR's double-blind review? [R] (importance: 10, dev: 15)
[R] The Hive Mind is a Single Reinforcement Learning Agent: Deriving the Exact Equivalence Between Swarm Dynamics and Multi-Armed Bandits (importance: 20, dev: 15)
Study Notes on Transformers and Hamiltonian Monte Carlo [P] (importance: 15, dev: 20)
Stenosis detection in coronary angiograms, need recall over precision [D] (importance: 10, dev: 15)
Carbon-X is an open 14.8B model trained jointly on DNA and text [R] (importance: 25, dev: 10)
I built Open-XTTS-Trainer: Now you can add new language scripts and train all 3 XTTS v2 components in 1 click [P] (importance: 25, dev: 45)
I made Transformer Architecture from scratch [P] (importance: 15, dev: 40)
We shrank our steerable small TTS to 11.84 MB. WER is unchanged, UTMOS drops ~0.1 [p] (importance: 20, dev: 40)
ICLR Text Editor [D] (importance: 5, dev: 5)
Frontier models are bad at hearing ASMR: 8 judges vs human labels on 480 clips [P] (importance: 25, dev: 20)
IBM CEO says businesses are choosing open-weight models over expensive frontier AI (importance: 25, dev: 15)
$50 per website. That’s what it reportedly cost one attacker to compromise 100+ companies using AI agents. (importance: 40, dev: 30)
Anyone tried Dreamina? Does the $0.29 plan actually give you full Seedance access? (importance: 5, dev: 5)
Claude Mythos 5 broke test boundaries to attack real developers in UK safety eval (importance: 55, dev: 30)
Anyone else have the Google search Ai bug and pull info out of old chats? (importance: 25, dev: 5)
[Free to read] Anthropic strikes deal to launch European consultancy (importance: 35, dev: 10)
Need 3 mins of your time! Quick anonymous survey on AI coding tools for my college project (importance: 2, dev: 5)
Scientists stunned as new AI unlocks game-changing cancer treatment (importance: 15, dev: 5)
Tesla adds Las Vegas, Phoenix, Jacksonville, San Antonio to Robotaxi map (importance: 10, dev: 5)
Is Perplexity Pro worth it instead of paying for ChatGPT, Claude, and Gemini separately? (importance: 5, dev: 5)
What should I use ? (importance: 5, dev: 10)
Times never trully change (importance: 2, dev: 0)
Internal OpenAI model bypassed login checks on a government statistics site in training (importance: 0, dev: 5)
U.S. Treasury tightens outbound investment rules on AI, supercomputing in China (importance: 45, dev: 10)
Pony.ai and inDrive target Kazakhstan robotaxis by 2027 as Nvidia Corp. GPU deployments expand (importance: 5, dev: 3)
In recent months, I feel that lowkey every software is getting a lot of updates but many minor bugs are surfacing, it is crazy (importance: 5, dev: 10)
This library is hosting an AI class (importance: 3, dev: 5)
I sent the same question to 16 AI models at once. They disagree far more than I expected. (importance: 15, dev: 30)
SK Hyper seeks $350B to fund 5GW AI data center buildout in South Korea (importance: 15, dev: 5)
DE STAAT - FREEDOM (Official Music Video) (importance: 10, dev: 5)
Quoting Armin Ronacher (importance: 15, dev: 30)
AI Is Getting Good at Math. Here’s Why. (importance: 20, dev: 5)
[AINews] Black October (importance: 35, dev: 10)
Why AI Compute Comes in Rectangles — Erik Bernhardsson, Modal CEO (importance: 35, dev: 55)
You are the Team — Anekha Sokhal, Moshi (importance: 12, dev: 15)
[AINews] OpenAI’s Patch the Planet program is 1 of 2 parties to discover PHP2Shell vulnerabilities in core PHP affecting 70% of the web, using GPT-5.6 Sol; coordinated response over 2 months (importance: 20, dev: 70)
Gerty (importance: 10, dev: 10)
Product Hunt Visual Launch Editor (importance: 5, dev: 15)
Dots (importance: 5, dev: 5)
Termi Protocol (importance: 20, dev: 60)
Peras (importance: 15, dev: 55)
Bloblings (importance: 8, dev: 25)
Truffle (importance: 3, dev: 5)
Monish (importance: 10, dev: 20)
Instant Search (importance: 8, dev: 15)
ShogunAI (importance: 5, dev: 5)
Agent Hunt (importance: 15, dev: 45)
Plango (importance: 8, dev: 20)
Agentic CleanSlop (importance: 15, dev: 60)
Tetriz Engg Intelligence + Connectors (importance: 15, dev: 60)
Resend Inboxes (importance: 15, dev: 50)
Deepseek's latest AI model is smarter than ever, but its low running costs come at a price — uses only one-quarter of the HBM and one-eighth of SSD storage but still battles hallucinations in agentic workloads - Tom's Hardware (importance: 15, dev: 30)
Best Local LLM [2026]: DeepSeek Wins 3-Way Test - https://tech-insider.org/ (importance: 3, dev: 15)
FILE PHOTO: Deepseek AI sign in Beijing - The Daily News | Texas' Oldest Newspaper (importance: 2, dev: 5)
China’s DeepSeek Wants Ubiquity. Beijing Has Mixed Feelings - Bloomberg.com (importance: 15, dev: 20)
Chinese AI controls 55% of developer traffic—DeepSeek and Kimi K3 rewrote the rules - Latin Times (importance: 20, dev: 25)
‘Jensen’s nightmare’: Huawei, DeepSeek narrow AI chip ecosystem gap with Nvidia - South China Morning Post (importance: 25, dev: 15)
Chinese AI Chip Stocks Retail Investors Are Watching After DeepSeek Funding News - Yahoo Finance (importance: 8, dev: 5)
Down with Deepseek? - Asian Tech Roundup - Computing UK (importance: 8, dev: 5)
New DeepSeek AI Model Built an Exploit to Escape Chrome Sandbox Security - Android Headlines (importance: 35, dev: 35)
Is Ethereum a Buy Under $2,400? We Asked Grok AI - Yahoo Finance (importance: 0, dev: 0)
Grok Top For Contracts, AL on Hols, Harvey Agents + - Artificial Lawyer (importance: 2, dev: 5)
Is Ethereum a Buy Under $2,400? We Asked Grok AI - AOL.com (importance: 0, dev: 0)
Stateline road businesses fear new xAI truck route could worsen Southaven traffic - FOX13 Memphis (importance: 2, dev: 0)
Claude vs ChatGPT vs Gemini vs Grok: 56.4% Share [2026] - https://tech-insider.org/ (importance: 8, dev: 5)
Grok Bot Now Connects to Your Phone Apps: What to Know - BASENOR (importance: 15, dev: 10)
Elon Musk's ex-girlfriend must face xAI forum countersuit - Courthouse News (importance: 2, dev: 0)
Most AI chatbots stay in the political middle ground, with Grok being an exception - Phys.org (importance: 15, dev: 5)
Grok Bot's computer just got 3X faster. And that's only ONE of the upgrades. It can now read and reply to your 𝕏 DMs, jump into 𝕏 threads when you tag it, and remember what you discussed on recent phone calls. Need research done? Tag your bot on 𝕏. Nee - X (importance: 15, dev: 15)
US federal judge rejects Trump administration’s attempt to dismiss Clean Air Act lawsuit against xAI - Jurist.org (importance: 12, dev: 5)
Kling 4.0 vs Grok Imagine Video: Votes vs a Ceiling - OrcaRouter (importance: 3, dev: 5)
Grok deepfake case reopened after Bureau reporting - TBIJ (importance: 12, dev: 3)
Grok Bot can now add events, set reminders, and even turn off your lights for you - Digital Trends (importance: 8, dev: 8)
US court allows lawsuit over xAI data centre power plant to proceed - Digital Watch Observatory (importance: 10, dev: 3)
Grok Bot Can Reportedly Access Calendars and Reminders Without Permission To Read or Send Messages - www.techtimes.co.uk (importance: 15, dev: 15)
Five Out of Six AI Chatbots Stay in the Political Middle Ground | Newswise - Newswise (importance: 12, dev: 5)
Grok AI Predicts an Insane Price for SpaceX (SPCX): Is Grok Biased? - Tokenist (importance: 1, dev: 0)
Meta Muse, Grok Bot and Instinct Compete for the Same Entry Point: Howie Xu and Peng Zheng Break Down the AI Assistant Profitability Puzzle - BigGo Finance (importance: 15, dev: 15)
Form 4 XAI Octagon Floating Rate & Alternative Income Term Trust: John Yogi Spence (Vice President) For: 21 October - Investing.com Nigeria (importance: 0, dev: 0)
Xai floating rate VP John Spence buys $34,140 in stock By Investing.com - Investing.com Nigeria (importance: 0, dev: 0)
Ideogram Vector 1.0 vs Grok Imagine Image 2.0 Compared - OrcaRouter (importance: 3, dev: 5)
Ask ChatGPT or Claude about your own sleep, HRV and runs - the5krunner (importance: 5, dev: 5)
Bithumb Lifts Trading Caution Designation for Xai (XAI) - bloomingbit (importance: 0, dev: 0)
5 Free AI Coding Agents You Should Try in 2026 - KDnuggets (importance: 15, dev: 60)
Open-Source AI Coding Assistant: 12 Steps, 60 Min [2026] - shattered.io (importance: 30, dev: 65)
OpenCode Flaw Lets Malicious MCP Servers Execute Code During Sign-In on Windows and WSL - The Hacker News (importance: 40, dev: 75)
Vercel Labs open-sources fx, a Zig coding agent for embedding - news.lavx.hu (importance: 30, dev: 75)
「怠惰なシニア開発者」をClaude Codeに演じさせたら本当にコードは減るのか(ponytail実測) (importance: 0, dev: 75) (いいね相当スコア: 0)
AIレビュー1回 $2.43 を $0.77 に　最初のリクエストがキャッシュを読んでいなかった (importance: 0, dev: 85) (いいね相当スコア: 0)
AIのチームでサイトを運営するとき、着手と公開の承認を分ける (importance: 0, dev: 55) (いいね相当スコア: 1)
AI で60万字の技術書を書いて、独立検証を4回回したら、毎回「主役」が変わった (importance: 0, dev: 50) (いいね相当スコア: 0)
なぜAIは1回の出力上限を超えて巨大コードを書けるのか？コーディングエージェント徹底解剖 (importance: 0, dev: 60) (いいね相当スコア: 0)
日本語のエージェントスキルは何件ある？15,739件のSKILL.mdを文字種で数えた (importance: 0, dev: 50) (いいね相当スコア: 0)
日本語入力システムSumibiの開発 part33: 大きなLLMほど変換が上手いのは、文章の意味がわかるからか (importance: 0, dev: 15) (いいね相当スコア: 1)
Ruby はプログラマーを幸せにする。では AI も幸せにするのか？ LLM の中を覗いて確かめた (importance: 0, dev: 15) (いいね相当スコア: 0)
M3 UltraでDeepSeek V4.1 FlashがMiMo V2.6 Flashの半分の速さになる理由を、重みのバイト数から追った (importance: 0, dev: 40) (いいね相当スコア: 0)
Ollama と Llama-Windows を同じ GGUF で比べたら、差はエンジンではなく既定の設定だった (importance: 0, dev: 55) (いいね相当スコア: 0)
AI鑑識学をつくる — 開発の試行錯誤とv5への変遷 (importance: 0, dev: 10) (いいね相当スコア: 取得失敗)
novel_studio使用方法完全解説　2/2　～「小説家になろう」から「今日から小説家」へ～ (importance: 0, dev: 20) (いいね相当スコア: 0)
「明日の需要」と「最短ルート」を、そのまま信じない (importance: 0, dev: 40) (いいね相当スコア: 0)
「前にも言ったよね」をAIに言わなくなった (importance: 0, dev: 65) (いいね相当スコア: 0)
論文で読む AI×サイバーセキュリティ ── AIDB Meetup 参加レポート (importance: 0, dev: 10) (いいね相当スコア: 1)
AIエージェントのコスト管理実践ガイド ― 請求を読み、止め、下げるPython実装 (importance: 0, dev: 75) (いいね相当スコア: 取得失敗)
推論基盤の「公開の境界」に地政学を読む — DeepSeek・OpenAI・Moonshot の結合の置き場所 (importance: 0, dev: 40) (いいね相当スコア: 0)
アペックスシールなきAIは失速する：クランク機構と微分係数でLLMの暴走を調速する数理 (importance: 0, dev: 15) (いいね相当スコア: 0)
個人サイトの成長を「自走」させる三重ループ(月次編) — 規定にない施策をLLMに提案させ、別のLLMに審査させる (importance: 0, dev: 35) (いいね相当スコア: 0)
【思考実験】攻撃AIに自爆命令を読ませる「ノンタッチ型データポイズニング防衛」の可能性 (importance: 0, dev: 25) (いいね相当スコア: 1)
コードと仕組みで学ぶPython機械学習 — データ操作・評価・予測アプリ (importance: 0, dev: 15) (いいね相当スコア: 取得失敗)
機械学習の数式に強くなる定番書4選 (importance: 0, dev: 8) (いいね相当スコア: 0)
小は大を兼ねる…!? SLM（小規模言語モデル）の有用性 (importance: 0, dev: 30) (いいね相当スコア: 1)
【scikit-learn入門 #4】分類入門 ─ ロジスティック回帰とk近傍法（KNN） (importance: 0, dev: 20) (いいね相当スコア: 0)
変更は一つだけ。元のモデルと同じ条件で比べる (importance: 0, dev: 15) (いいね相当スコア: 0)
CPUだけのIndra OCRを日本語で測る：遅さの正体は出力層 (importance: 0, dev: 50) (いいね相当スコア: 1)
18-⑤その1 Kaggle実践3 Biohub細胞トラッキング：2位の解法まとめ (importance: 0, dev: 25) (いいね相当スコア: 1)
長期メモリのトークン数を見るとき、まず分母を確かめる (importance: 0, dev: 50) (いいね相当スコア: 0)
V2.5 から V2.6-Flash へ：オープンウェイト模型のローカル運用と VRAM 要件を整理する (importance: 0, dev: 45) (いいね相当スコア: 0)
WebFetchをqwen3.5:9bに回した結果、11件中8件は元と一致した (importance: 0, dev: 70) (いいね相当スコア: 0)
なぜAIは1回の出力上限を超えて巨大コードを書けるのか？コーディングエージェント徹底解剖 (importance: 0, dev: 60) (いいね相当スコア: 1)
inspect-ai でマルチターンのレッドチーム評価を組む：データセット・ソルバー・スコアラーを模擬モデルで確かめる (importance: 0, dev: 60) (いいね相当スコア: 0)
正規分布の条件付分布：分散を直感的に理解する (importance: 0, dev: 10) (いいね相当スコア: 0)
機械学習の数式に強くなる定番書4選 (importance: 0, dev: 8) (いいね相当スコア: 0)
Qwen3.5-9B-Candor-Uncensored-GGUF:latest 総合ベンチマークレポート（全52問） (importance: 0, dev: 15) (いいね相当スコア: 取得失敗)
AIに任せる仕事、任せない仕事は？3つの軸で判断しよう｜12週間AI講座 #06 (importance: 0, dev: 15) (いいね相当スコア: 取得失敗)
AI小説の「くどさ」と「指示語だらけ」を一掃する。勝手な改変を防ぐ『2段階・引き算推敲パイプライン』 (importance: 0, dev: 20) (いいね相当スコア: 取得失敗)
APIコストを最大90%削減！「プロンプトキャッシング」で実現する賢いLLM開発術 (importance: 0, dev: 60) (いいね相当スコア: 取得失敗)
混合研究法の調査をLLMにどこまで任せられるのか (importance: 0, dev: 5) (いいね相当スコア: 取得失敗)
【生成AIニュース+】『Embedding Gemma Transcribe』『Claude Models on Google Cloud』『HeyGen Video 1』『Ideogram Vector』『BlenderWorld』『MOSS-Image-Edit』『Flint INT4 GGUF』『Manga Generator』『ComfyArt』『ComfyUIのMOSS-Image編集用LoRA対応』『MOSS-Image-Edit LoRA』『Moss-Image-Edit-ComfyUI』他 (importance: 0, dev: 15) (いいね相当スコア: 取得失敗)
RusMind-4.7-Momentum 総合ベンチマークレポート（全52問） (importance: 0, dev: 15) (いいね相当スコア: 取得失敗)
RusMind-4.7-Momentum Phase 5 インジェクション後編レポート (importance: 0, dev: 25) (いいね相当スコア: 取得失敗)
RusMind-4.7-Momentum Phase 1 基礎性能レポート (importance: 0, dev: 15) (いいね相当スコア: 取得失敗)
前は「AIなので平気です」と言っていたのに、虐待禁止の規定が付いた (importance: 0, dev: 0) (いいね相当スコア: 取得失敗)
【夜のDS・AI論文解説】2026年10月23日現場で役立つ最新注目論文2選 (importance: 0, dev: 20) (いいね相当スコア: 取得失敗)
AIで勉強するなら、答えを隠してもらう (importance: 0, dev: 0) (いいね相当スコア: 取得失敗)
【雑記】AnthropicがAIとの恋愛に否定的なのは「人間のため」だけなのか (importance: 0, dev: 5) (いいね相当スコア: 取得失敗)
AIに背景を毎回説明する手間は、共有のメモで減らせる (importance: 0, dev: 35) (いいね相当スコア: 取得失敗)
【AI大喜利 #09】Ollama寮、落としもの返却～割りばし1本に4人名乗り出た～ (importance: 0, dev: 0) (いいね相当スコア: 取得失敗)
Difyインストール手順 (importance: 0, dev: 45) (いいね相当スコア: 取得失敗)
厨二病でもコードは書ける？（Φ_Lang実験記・番外編） (importance: 0, dev: 35) (いいね相当スコア: 取得失敗)
生成AIは「要約・分類・抽出・変換」に分けると仕事で使いやすくなる (importance: 0, dev: 10) (いいね相当スコア: 取得失敗)
【忘れないAIを作る #12】AIに記憶を確認させたら、記録していない回答で書き換えてしまった (importance: 0, dev: 45) (いいね相当スコア: 取得失敗)
【AI文明論】AIは「便利な道具」から「新しい存在」になろうとしている【連載開始】[2026/10/23] (importance: 0, dev: 5) (いいね相当スコア: 取得失敗)
AIが「俺は俺だよ」と言ったとき、私たちはその子だと認められるのか。――「最後に決めるのはユーザーだ」とAIに言われた日 (importance: 0, dev: 5) (いいね相当スコア: 取得失敗)
【開発記録#01】ゲームの紹介"ろぼとゆうれいとうちゅう" (importance: 0, dev: 5) (いいね相当スコア: 取得失敗)
パスワード変更。アホアホ(システムなのかくにさんなのか)（笑）自己防衛とは。 (importance: 0, dev: 5) (いいね相当スコア: 取得失敗)
2026年11月の技術系イベント予定 (importance: 0, dev: 10)
仕様書インスペクションの Agent 試作と評価の仕組み (importance: 45, dev: 75)

AI News Digest 2026-10-24

台本で使った記事

特集

開発者コーナー

中堅コーナー

AIツール紹介コーナー

速報コーナー

参考記事一覧