MagicBeanStalk: Escaping the RLVR Zero Advantage Trap with Recursive Data Synthesis

Ximing Lu et al.

#arXiv #cs.AI #cs.LG

Know What You Don't Know: Agentic Data Synthesis by Evolving Generation Space for Closed-Loop Learning

Wei Yang et al.

#arXiv #cs.AI #cs.MA

New arXiv work combines Monte Carlo sampling with regression to approximate Shapley, Banzhaf, and other probabilistic values more flexibly, even using tree-based models like XGBoost for unbiased estimates.

#AI #MachineLearning #ExplainableAI #ArXiv
https://arxiv.org/abs/2506.11849

KnowForge separates document QA into a Locator agent, a deep Reader, and a typed evidence pool, instead of letting one model handle both retrieval and reading on its own. Reported F1 of 47.5 on M3DocVQA suggests the role-specialized…

#arXiv #AIagents #RAG #devtools
https://arxiv.org/abs/2610.13128

New arXiv paper systematically re-evaluates regression objectives for tabular deep learning, finding many proposed alternatives offer limited gains over tuned MSE, while quantile loss shows consistent improvements. The…

#TabularData #DeepLearning #Regression #ArXiv
https://arxiv.org/abs/2610.13909

Beyond Domain-Routed Supervision: Cross-Teacher Guarding for Multi-Teacher On-Policy Distillation

Shi-Yu Tian et al.

#arXiv #cs.AI

Masked Diffusion Language Agents for Tool-Integrated Chemical Reasoning

Mengdi Liu et al.

#arXiv #cs.AI #cs.LG

Travel Back To Move Forward: Retrospective Exploration and Evolving for Long Horizon Agents

Ziang Ye et al.

#arXiv #cs.AI #cs.CL

Provenance-Aware Belief Updating in Human-AI Feedback Loops

Tianyi Huang et al.

#arXiv #cs.AI

DeFiRank: Federated Pairwise Ranking for Governance in Decentralized Financial Networks

Zahra Batool et al.

#arXiv #cs.AI

VEPA: Verified Evolving Formal Policy for Lifelong LLM Agent Guardrails

Duy Nguyen et al.

#arXiv #cs.AI #cs.CL #cs.CR #cs.LG

CLEAT: Learning Value-Guided Clarification for Interactive Agents

Jiatan Huang et al.

#arXiv #cs.AI

Equal-Budget Monte Carlo Tree Search versus Randomized-Restart List Scheduling for DAG Workflows: A Controlled Equal-Budget Computational Study

Yuyu Zhang

#arXiv #cs.AI #cs.DC #cs.LG

New arXiv paper models how AI shopping assistants shape retail assortments, finding that accounting for AI mediation can materially change optimal catalogs and revenue relative to standard choice models.

#retailAI #assortmentPlanning #choiceModels #arxiv
https://arxiv.org/abs/2610.12917

A new arXiv paper proposes multi-agent session alternation for long-horizon machine learning engineering, framing it as a context-reset semi-Markov decision process to address "long-horizon deceleration" and…

#MLEAgents #LLMAgents #ContextEngineering #arxiv
https://arxiv.org/abs/2610.13494

🔭 Radar arXiv del Día | Frontera Científica

¿Podemos atrapar a una IA mintiendo antes de que lo verbalice? Nuevas sondas detectan sabotaje y engaño oculto en modelos de frontera.

📄 "Caught in the Act: Probe...
👉 Ver en Radar LOGIDIAS
#arXiv #Ciencia #CienciaÑ #SciSky #Física #Física

🔭 Daily arXiv Radar | Scientific Frontier

Can we catch AI deception before it speaks? New internal probes successfully detect hidden sabotage and unverbalized intent!

📄 "Caught in the Act: Probes Effectively Detec...
👉 Read on LOGIDIAS Radar
#arXiv #Science #SciSky #Physics #Physics

Beyond Domain-Routed Supervision: Cross-Teacher Guarding for Multi-Teacher On-Policy Distillation

Shi-Yu Tian et al.

#arXiv #cs.AI

Masked Diffusion Language Agents for Tool-Integrated Chemical Reasoning

Mengdi Liu et al.

#arXiv #cs.AI #cs.LG

Travel Back To Move Forward: Retrospective Exploration and Evolving for Long Horizon Agents

Ziang Ye et al.

#arXiv #cs.AI #cs.CL