arXiv AI By Renuka Oladri, Niveda Jawahar, Abdirisak Mohamed

Token Budget Saturation and Mechanistic Early Detection of Reasoning Non-Convergence in Chain-of-Thought Models

Read the original on arXiv AI →

arXiv:2607. 21433v1 Announce Type: cross Abstract: Chain-of-thought reasoning models such as DeepSeek-R1-Distill-Qwen-7B exhibit a bimodal convergence pattern: generations either terminate within a token budget (converged) or exhaust it without reaching a conclusion (non-converged).

Summary generated by The Flow from the publisher's feed. The full article lives at arXiv AI.