AI News

WorldBench: Evaluating LLMs on Three.js Voxel World Generation (arxiv.org)

When AI Finds Hidden Messages, Does It Report? (arxiv.org)

TestJack: Should you trust the results in coding benchmarks? Agentic Coding Benchmarks Auditing via Evaluator Evolution (arxiv.org)

Code Understanding is a Bottleneck for Coding Agents (arxiv.org)

From Investigation Failures to Reliable SOC Agents: Understanding and Improving LLM-Based Alert Triage (arxiv.org)

Beyond Type-checking: Towards Holistic Evaluation of Formal Specification Generation (arxiv.org)

Certified Corruption Budgets: Anytime-Valid Leaderboard Claims under Adaptive Rigging (arxiv.org)

Unlocking the Regulatory Genome by ARGUS: An Evidence-Constrained Agentic Framework for Interpreting Single Nucleotide Variants (arxiv.org)

On linearity or non-linearity in machine learning for quantum chaotic dynamics (arxiv.org)

Safe at One Loop, Risky at Another: Aligning Safety Across Recurrent Depths in Looped Language Models (arxiv.org)

MRCert: Towards Post-deployment Patch Robustness Certification for Adversarially Patched Samples via Type-specific Masking (arxiv.org)

GaussianBench: Physics-Fidelity Evaluation for Gaussian Scene Representations (arxiv.org)

Could LLM Watermark Detection be Public? (arxiv.org)

A Survey on LLM-Integrated Hardware Design Verification (arxiv.org)

SLVR: Structured Latent Visual Reasoning via Human-like Reasoning Flows (arxiv.org)

Strategic Governance of AI Models in Earth Science (arxiv.org)

From Pixels, Without Pre-training: Joint Generative and Self-Supervised Representation Learning in One Model (arxiv.org)

Bi-FORK: Generative Modeling of High-Dimensional Bifurcating Systems (arxiv.org)

Caught in the Act: Probes Effectively Detect Sabotage and Catch Unverbalized Deception (arxiv.org)

Rounding in Preconditioner Space: Redesigning 4-bit AdamW Optimizer-State Quantization (arxiv.org)

A Unified Bellman Operator for Safety-Critical Reinforcement Learning (arxiv.org)

Learning Kilometer-Scale Weather Prediction with Global-Regional Alignment (arxiv.org)

Interpretable Memory Models for Spaced Repetition (arxiv.org)

On the estimation and validity of AI time horizons---a statistical look at the METR plot (arxiv.org)

BrickBench: Evaluating Agentic Brick Design (arxiv.org)

Ecology of AI Agents: Collaboration Creates a Population Threshold for Takeoff (arxiv.org)

Searching for "Harmful Refusal": A Psychometric Audit of an AI Safety Benchmark (arxiv.org)

HRIL: Learning Multimodal Synergy via Higher-Order Tensor Modeling (arxiv.org)

SpectralCache: Accelerating Diffusion-Based World Models via Spectral Feature Caching (arxiv.org)

Prospective Prediction of OOD Degradation from Source-Side Training Dynamics (arxiv.org)

Long Text to Predictive Features: LLM-Guided Blockwise Feature Engineering via Executable Program Search (arxiv.org)

Marformer: A Transformer for Predicting Missing Data Distributions (arxiv.org)

SplitJEPA: Learning Invariant and Variant Latent Worlds without Reconstruction (arxiv.org)

GeoReform: Reflective Formalization Evolution for Multimodal Geometry Problem Solving (arxiv.org)

OnTrack: Real-Time Monitoring and Intervention in LLM Agent Trajectories via Streaming Structure-Aware Optimal Transport (arxiv.org)

HANS: A Handwritten Answer Sheet Dataset for Noisy Hybrid Document Parsing (arxiv.org)

Cited but Not Consulted: A Counterfactual Audit of Legal Chain-of-Thought Faithfulness (arxiv.org)

Accurate but Not Humble: Evaluating Epistemic Humility in LLM Agents under Knowledge Conflict (arxiv.org)

Overcoming Prior Barriers: Supervised Fine-Tuning under Long-Tail Distribution (arxiv.org)

Can AI Agents Learn Their Way to the Top? Evaluating Heuristic Learning in a Long-Running Game Agent Competition (arxiv.org)

Prior or Feedback? What an LLM Uses When Adapting Neural Operators (arxiv.org)

SparseDecoding: Decoding-Aware Pruning for Accurate and Efficient LLM Inference (arxiv.org)

AdaptLSTM: Efficient Adaptive Online Learning for Cloud Workload Forecasting under Distribution Drift (arxiv.org)

Batch Before You Lift: Scalable Topological Deep Learning on Large Graphs (arxiv.org)

Training on the Future: A Delay-Aware Audit of Test-Time Adaptation for Time-Series Forecasting (arxiv.org)

DataSense-Bench: The First Step Toward an AI Scientist (arxiv.org)

Looking Inside LLMs: Small-World Connectivity as a Signature of Reasoning Performance (arxiv.org)

Learning Probabilistic Logic Programs with Functional Gradient Guided Language Models (arxiv.org)

One Word Opens the Gate: The Option-Channel Attack on Typed Decision Models as Agent Guardrails (arxiv.org)

When Has a Bayesian Neural Network Sampled Enough? Adaptive Inference Time with Statistical Guarantees (arxiv.org)