Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | https://arxiv.org/abs/2509.14004 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866910232745082880 |
|---|---|
| author | Mao, Minjia Yin, Bowen Zhu, Yu Fang, Xiao |
| author_facet | Mao, Minjia Yin, Bowen Zhu, Yu Fang, Xiao |
| contents | Reasoning large language models (LLMs) have demonstrated superior capacities in solving complicated problems by generating long chain-of-thoughts (CoT), but such a lengthy CoT incurs high inference costs. Previous methods on inference-stage efficient reasoning either require white-box models to monitor the reasoning process or are not reliable through direct prompting. In response, we introduce ES-CoT, an inference-time method that shortens CoT generation by detecting answer convergence and stopping early with almost no performance loss. When observing a linguistic marker (such as "wait") in the reasoning process, we prompt the LLM to output its current final answer, denoted as a step answer. We then track the run length of consecutive identical step answers as a measure of answer convergence. We show both empirically and theoretically that step answers steadily converge to the final answer, and large run-length jumps reliably mark this convergence. Experiments on six reasoning datasets across three LLMs show that ES-CoT reduces the number of inference tokens by 16.08% on average while maintaining accuracy comparable to standard CoT. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_14004 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Early Stopping Chain-of-thoughts in Large Language Models Mao, Minjia Yin, Bowen Zhu, Yu Fang, Xiao Computation and Language Reasoning large language models (LLMs) have demonstrated superior capacities in solving complicated problems by generating long chain-of-thoughts (CoT), but such a lengthy CoT incurs high inference costs. Previous methods on inference-stage efficient reasoning either require white-box models to monitor the reasoning process or are not reliable through direct prompting. In response, we introduce ES-CoT, an inference-time method that shortens CoT generation by detecting answer convergence and stopping early with almost no performance loss. When observing a linguistic marker (such as "wait") in the reasoning process, we prompt the LLM to output its current final answer, denoted as a step answer. We then track the run length of consecutive identical step answers as a measure of answer convergence. We show both empirically and theoretically that step answers steadily converge to the final answer, and large run-length jumps reliably mark this convergence. Experiments on six reasoning datasets across three LLMs show that ES-CoT reduces the number of inference tokens by 16.08% on average while maintaining accuracy comparable to standard CoT. |
| title | Early Stopping Chain-of-thoughts in Large Language Models |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2509.14004 |