AI News

Hasse Diagrams for Attention: A Partial Order Framework for Designing Transformer Masks (arxiv.org)

Integrating Out, Twice:The Open-System Case That Neural-Network Ensemble Theory Is Missing (arxiv.org)

Culturally-Aware AI for Cross-Boundary Community Learning: Undergraduate Innovation at the Intersection of Computation and Design (arxiv.org)

A History-Aware Visually Grounded Critic for Computer Use Agents (arxiv.org)

CIAware-Bench: Benchmarking Control Intervention Awareness Across Frontier LLMs (arxiv.org)

What Fits (Into Few Tokens) Doesn't Overfit: Compression and Generalization in ML Research Agents (arxiv.org)

Workflow-GYM: Towards Long-Horizon Evaluation of Computer-use Agentic tasks in Real-World Professional Fields (arxiv.org)

Convergence of Monte Carlo Optimistic Policy Iteration: Beyond Uniform State-Action Updates (arxiv.org)

Superficial Beliefs in LLM Decision-Making (arxiv.org)

Structure from Reasoning, Numbers from Search: On-Premise Open LLMs as Structural Priors for Coupled MIMO Controller Tuning (arxiv.org)

Null-Space Constrained Low-Rank Adaptation for Response-Specified Large Language Model Unlearning (arxiv.org)

Bellman-Taylor Score Decoding for Markov Decision Processes with State-Dependent Feasible Action Sets (arxiv.org)

Mind the Gap: Can Frontier LLMs Pass a Standardized Office Proficiency Exam? (arxiv.org)

MMClima: A Framework for Multimodal Climate Science Data and Evaluation (arxiv.org)

One Lens, Many Worlds : A Capability-Typed Interface for World-Model Interpretability (arxiv.org)

nCMD: Benign-Anchored Feature Selection for Imbalanced Network Intrusion Detection (arxiv.org)

When RL Fails after SFT: Rejuvenating Model Plasticity for Robust SFT-to-RL Handoff (arxiv.org)

Between Amnesia and Chaos: A Memory Stability Expressivity Trilemma for Trainable Dissipative Oscillator Networks (arxiv.org)

Unsupervised Deep Learning for Limited-Angle STEM-EDX Tomography -- Application to 3D Chemical Analysis of Phase-Change Memory Devices (arxiv.org)

Architect-Ant: Editable Automatic Furnishing of Architectural Floor Plans (arxiv.org)

Recalling Too Well: Sycophancy Evaluation and Mitigation in Memory-Augmented Models (arxiv.org)

WorldKernel: A World Model is the Coupling Kernel of Admissible Possible Worlds (arxiv.org)

Frontier Coding Agents Use Metaprogramming to Adapt to Unfamiliar Programming Languages (arxiv.org)

Trading Utility for Dynamic Fairness in Multiple Resource Division with Sequential Demand (arxiv.org)

Bittensor Agent Arenas as a Trajectory Primitive: Distilling a Shopping Agent from ShoppingBench Subnet Traces (arxiv.org)

Forward-Only Convolutional Neural Networks with Learnable Channel-Class Assignment (arxiv.org)

Trainable Smooth-Rotation Transforms with Learned Channel Scales for LLM Quantization (arxiv.org)

Sample Where You Struggle: Sharpening Base Model Reasoning via Entropy-Guided Power Sampling (arxiv.org)

Sigma-Branch: Hierarchical Single-Path Network Reconstruction for Dynamic Inference with Reduced Active Parameters (arxiv.org)

Importance-Aware Scheduling for High-Dimensional Hyperparameter Optimization (arxiv.org)

Role-Agent: Bootstrapping LLM Agents via Dual-Role Evolution (arxiv.org)

Large-scale semantic mapping of learner agency and autonomy reveals what measurement and generative AI research overlook (arxiv.org)

Do VLMs Reason Like Engineers? A Benchmark and a Stage-wise Evaluation (arxiv.org)

Moonshine: An Autonomous Mathematical Research Agent Centered on Conjecture Generation (arxiv.org)

Interpreting and Steering a Text-to-Speech Language Model with Sparse Autoencoders (arxiv.org)

Mix, Don't Pick: Why Synthetic Corpus Composition Matters for Time Series Foundation Model Pretraining (arxiv.org)

LongMoE: Longitudinal Multimodal Learning via Trajectory-Aware Mixture-of-Experts (arxiv.org)

TRAPS: Therapeutic Response Analysis via Pathway-informed Stratification (arxiv.org)

A Navigable Manifold of Hypothesized Consciousness-Spectrum States in Language Model Representations (arxiv.org)

Optimality of FSQ Tokens for Continuous Diffusion for Categorical Data with Application to Text-to-Speech (arxiv.org)

Evaluating Research-Level Math Proofs via Strict Step-Level Verification (arxiv.org)

READER: Robust Evidence-based Authorship Decoding via Extracted Representations (arxiv.org)

More Human or More AI? Visualizing Human-AI Collaboration Disclosures in Journalistic News Production (arxiv.org)

Accelerating NeurASP with vectorization and caching (arxiv.org)

$\tau$-Rec: A Verifiable Benchmark for Agentic Recommender Systems (arxiv.org)

AutoPDE: Reliable Agentic PDE Solving via Explicitly Represented Solver Strategies (arxiv.org)

The Arbiter Agent: Continually Monitoring Multi-Agent Conversations to Detect Emergent Misalignment (arxiv.org)

When the Chain of Thought Knows Better: Failure Modes in Multi-Turn Reasoning Models (arxiv.org)

Learning What to Remember: Observability-Safe Memory Retention via Constrained Optimization for Long-Horizon Language Agents (arxiv.org)

LMT: A Bayesian Framework for Causal Discovery from Textual Alarm Records in Manufacturing Systems (arxiv.org)