New arXiv work uses probability signatures to explain how two-layer networks learn modular addition, linking gradient dynamics to discrete Fourier structure and clarifying early-loss behavior under label noise.

#MachineLearning #DeepLearning #Interpretability #ArXiv
https://arxiv.org/abs/2610.11833

🔭 Daily arXiv Radar | Scientific Frontier

Can we listen to the edge of a black hole? New gravitational wave data probes near-horizon reflectivity like never before.

📄 "Listening to the Horizon: Probing Near-Horizo...
👉 Read on LOGIDIAS Radar
#arXiv #Science #SciSky #Physics #Physics

GeoReform: Reflective Formalization Evolution for Multimodal Geometry Problem Solving

Jialu Wang et al.

#arXiv #cs.AI

A Structural Theory of Cognitive Representation and Problem Solving,Contexts, Invariance, and the Knowledge Space

Antal Jakov\'ac, Andr\'as Telcs

#arXiv #cs.AI

Use and Disuse: Intent-Structured Experience Consolidation for Memory and Learning in LLM Agents

Xiangyi Zeng et al.

#arXiv #cs.AI

Complexity of Grounded Semantics and Preferred Semantics in Finitary Argumentation Frameworks

Jinfan Xu, Jieting Luo

#arXiv #cs.AI #cs.LO

An Interpretable Approach to PDE Solution Discovery via Structural Experience Distillation

Yunpeng Gong et al.

#arXiv #cs.AI

MetaOPD: Meta-Learned Token Weighting for On-Policy Distillation

Zipeng Wang et al.

#arXiv #cs.AI

Where Draft Trees Lose Target Mass: Exit-Guided Speculative Decoding

Shijing Hu et al.

#arXiv #cs.AI

Error-Propagation Modeling for Failure Attribution in LLM-Based Multi-Agent Systems

Jiaqi Liao et al.

#arXiv #cs.AI

Cognition-Oriented Emotion Tracing from Causes to Consequences in Real-World Social Scenes

Hao Li et al.

#arXiv #cs.AI

Writing for the Reviewer: Defensive Writing in GPT Models

Junchi Liao

#arXiv #cs.AI

TokenBank: Financial Infrastructure for AI Services

Cary Chang, Jialin Zhou

#arXiv #cs.AI #cs.CE

On the Clock: Towards Punctual and Productive Time-Budgeted AI Agents

Aaron Wang et al.

#arXiv #cs.AI #cs.LG

New arXiv paper proposes a 3D framework characterizing AI puzzle-solving methods via MDP formalism, autonomy, and computational cost, comparing Neurosolver, FBRL, and an LLM-based approach. Useful lens for evaluating…

#AI #ReinforcementLearning #LLMs #Arxiv
https://arxiv.org/abs/2610.11696

Study contracts bind declared experimental choices to execution evidence, separating contract-relative verification from scientific truth in AI research. A diagnostic checker reliably flagged registered mutations,…

#AIresearch #DataInfrastructure #AIAgents #Arxiv
https://arxiv.org/abs/2610.11754

Searching for "Harmful Refusal": A Psychometric Audit of an AI Safety Benchmark

Christopher M. Stewart et al.

#arXiv #cs.AI

HANS: A Handwritten Answer Sheet Dataset for Noisy Hybrid Document Parsing

Xiazhen Wu et al.

#arXiv #cs.AI #cs.CV

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

Kaisen Yang et al.

#arXiv #cs.AI #cs.CL

When Has a Bayesian Neural Network Sampled Enough? Adaptive Inference Time with Statistical Guarantees

Fabian Denoodt, Sibylle Hess

#arXiv #cs.AI