ThinkReset: Learnable Intermediate Interface Construction for Bounded-Context Long-Horizon Reasoning
Fei Ding, Yongkang Zhang, Runhao Liu, Yuhao Liao, Zijian Zeng
- Digest date
- 2026-08-03
- Submitted
- 2026-05-26
- arXiv ID
- 2607.28642
Summary
ThinkReset identifies that the real bottleneck in bounded-context long-horizon reasoning isn't trajectory compression but the lack of a reusable intermediate interface to replace discarded history. It also diagnoses a failure mode in outcome-reward-driven long-chain RL where models are incentivized to guess prematurely near context exhaustion instead of continuing to reason, and fixes both by explicitly writing/resetting interfaces and optimizing post-reset continuation success.
Abstract
Long chain-of-thought reasoning improves performance on complex problems, but it also introduces redundancy accumulation, context overflow, and error anchoring. We argue that under bounded context windows, the core bottleneck is not trajectory compression or test-time control, but the absence of a reusable intermediate interface that can replace discarded history and support continued solving. We further identify a key failure mode of outcome-reward-driven long-chain reinforcement learning: when the model has not solved the task before the window is nearly exhausted, the final-answer reward encourages premature guessing rather than continued careful reasoning. We propose ThinkReset, a text-space instantiation of this view. ThinkReset explicitly constructs reusable intermediate interfaces through interface writeback and reset, and directly optimizes post-reset continuation success. Across multiple long-horizon reasoning benchmarks, this perspective consistently improves success rates under fixed context windows.