ReProbe: Efficient Test-Time Scaling of Multi-Step Reasoning by Probing Internal States of Large Language Models
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914501286166528 |
|---|---|
| author | Ni, Jingwei Fadeeva, Ekaterina Wu, Tianyi Akhtar, Mubashara Zhang, Jiaheng Ash, Elliott Leippold, Markus Baldwin, Timothy Ng, See-Kiong Shelmanov, Artem Sachan, Mrinmaya |
| author_facet | Ni, Jingwei Fadeeva, Ekaterina Wu, Tianyi Akhtar, Mubashara Zhang, Jiaheng Ash, Elliott Leippold, Markus Baldwin, Timothy Ng, See-Kiong Shelmanov, Artem Sachan, Mrinmaya |
| contents | LLMs can solve complex tasks by generating long, multi-step reasoning chains. Test-time scaling (TTS) can further improve performance by sampling multiple variants of intermediate reasoning steps, verifying their correctness, and selecting the best steps for continuation. However, existing verification approaches, such as Process Reward Models (PRMs), are computationally expensive and require large-scale human or model-generated annotations. We propose a lightweight alternative for step-level reasoning verification based on probing the internal states of LLMs. We train a transformer-based probe that uses the internal states of a frozen LLM to estimate the credibility of its reasoning steps during generation. Annotation can be provided either by a larger LLM (e.g., DeepSeek-R1) or in a self-supervised manner by the original model itself. The probes are lightweight, containing fewer than 10M parameters. Across multiple domains, including mathematics, planning, and general knowledge question answering, our probes match or exceed the performance of PRMs that are up to 810x larger. These results suggest that LLM internal states encode confidence in their reasoning processes and can serve as reliable signals for step verification, offering a promising path toward scalable, generalizable TTS and more introspective LLMs. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_06209 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | ReProbe: Efficient Test-Time Scaling of Multi-Step Reasoning by Probing Internal States of Large Language Models Ni, Jingwei Fadeeva, Ekaterina Wu, Tianyi Akhtar, Mubashara Zhang, Jiaheng Ash, Elliott Leippold, Markus Baldwin, Timothy Ng, See-Kiong Shelmanov, Artem Sachan, Mrinmaya Artificial Intelligence Computation and Language LLMs can solve complex tasks by generating long, multi-step reasoning chains. Test-time scaling (TTS) can further improve performance by sampling multiple variants of intermediate reasoning steps, verifying their correctness, and selecting the best steps for continuation. However, existing verification approaches, such as Process Reward Models (PRMs), are computationally expensive and require large-scale human or model-generated annotations. We propose a lightweight alternative for step-level reasoning verification based on probing the internal states of LLMs. We train a transformer-based probe that uses the internal states of a frozen LLM to estimate the credibility of its reasoning steps during generation. Annotation can be provided either by a larger LLM (e.g., DeepSeek-R1) or in a self-supervised manner by the original model itself. The probes are lightweight, containing fewer than 10M parameters. Across multiple domains, including mathematics, planning, and general knowledge question answering, our probes match or exceed the performance of PRMs that are up to 810x larger. These results suggest that LLM internal states encode confidence in their reasoning processes and can serve as reliable signals for step verification, offering a promising path toward scalable, generalizable TTS and more introspective LLMs. |
| title | ReProbe: Efficient Test-Time Scaling of Multi-Step Reasoning by Probing Internal States of Large Language Models |
| topic | Artificial Intelligence Computation and Language |
| url | https://arxiv.org/abs/2511.06209 |