AI Papers - 2026-09-21

DailyArxiv - AI Research Podcast · Sep 21, 2026 · 16m

Today's papers: - CogGym: Towards Large-Scale Comparative Evaluation of Human and Machine Cognition: https://arxiv.org/abs/2609.21259v1 - The Communication Bottleneck: A Round-Trip Study of Tree-Structured Expression Serialization in Language Models: https://arxiv.org/abs/2609.21509v1 - Self Improvement via Fast Tree-search: https://arxiv.org/abs/2609.19526v1 - SWE-Proof: Can Language Models Resolve Real-World Issues with Machine-Checked Proofs?: https://arxiv.org/abs/2609.21190v1 - Quantifying Overclaiming Propensity in Frontier LLM Agents: https://arxiv.org/abs/2609.20812v1 This podcast is…