arXiv cs.AI· Tianxiang Chen, Jingyuan Zhou, Weiming Yang, Kaidi Yang·· 10 小时前AI 评分31
重新审视固定视野潜在推理:自适应调整潜在 CoT 推理长度
Rethinking Fixed-Horizon Latent Reasoning
AI 导读
研究发现,将潜在 CoT 推理扩展到预设的完整视野反而可能降低推理性能,因为后期计算主要在重组已有候选并削弱正确候选的支持。为此提出一种轻量方法,在推理时估计剩余推理进展并自适应调整潜在视野,无需重新训练 LLM。在多个算术基准和代表性潜在 CoT 推理器上,该方法将推理准确率最高提升 7.2 个百分点,同时减少最多 29.0% 的潜在计算量。
正文 · 原文
Abstract:Existing latent reasoners typically run for a preset latent horizon during inference. Intuitively, a longer horizon should provide more opportunity for reasoning. However, in this work, we find that extending latent Chain-of-Thought (CoT) to the full preset horizon may in fact degrade reasoning performance. We observe that latent CoT evolves through distinct functional stages, with earlier computation forming frontier reasoning states, while later computation mainly reorganizes existing candidates and can weaken gold candidate support. These findings motivate a rethink of fixed-horizon latent inference. To address this, we characterize the reasoning benefits of latent CoT by realized and remaining constructive progress. Empirically, we find that higher realized constructive progress corresponds to higher current reasoning accuracy, while more remaining progress corresponds to larger future gains from continued latent CoT. To this end, we introduce a lightweight method that estimates the remaining reasoning progress and adaptively adjusts the latent horizon during inference. Across representative latent CoT reasoners and multiple arithmetic benchmarks, our method improves reasoning accuracy by up to 7.2 percentage points while reducing latent computation by up to 29.0%, without retraining the LLM.
| Comments: | 21 pages, 4 figures |
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.24001 [cs.LG] |
| (or arXiv:2611.24001v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2611.24001 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Tianxiang Chen [view email]
[v1]
Wed, 18 Nov 2026 07:49:12 UTC (461 KB)
来源:arXiv cs.AI · arxiv.org