BrowseConf: Confidence-Guided Test-Time Scaling for Web Agents
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_ | 1866911236015259648 |
|---|---|
| author | Ou, Litu Li, Kuan Yin, Huifeng Zhang, Liwen Zhang, Zhongwang Wu, Xixi Ye, Rui Qiao, Zile Xie, Pengjun Zhou, Jingren Jiang, Yong |
| author_facet | Ou, Litu Li, Kuan Yin, Huifeng Zhang, Liwen Zhang, Zhongwang Wu, Xixi Ye, Rui Qiao, Zile Xie, Pengjun Zhou, Jingren Jiang, Yong |
| contents | Confidence in LLMs is a useful indicator of model uncertainty and answer reliability. Existing work mainly focused on single-turn scenarios, while research on confidence in complex multi-turn interactions is limited. In this paper, we investigate whether LLM-based search agents have the ability to communicate their own confidence through verbalized confidence scores after long sequences of actions, a significantly more challenging task compared to outputting confidence in a single interaction. Experimenting on open-source agentic models, we first find that models exhibit much higher task accuracy at high confidence while having near-zero accuracy when confidence is low. Based on this observation, we propose Test-Time Scaling (TTS) methods that use confidence scores to determine answer quality, encourage the model to try again until reaching a satisfactory confidence level. Results show that our proposed methods significantly reduce token consumption while demonstrating competitive performance compared to baseline fixed budget TTS methods. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_23458 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | BrowseConf: Confidence-Guided Test-Time Scaling for Web Agents Ou, Litu Li, Kuan Yin, Huifeng Zhang, Liwen Zhang, Zhongwang Wu, Xixi Ye, Rui Qiao, Zile Xie, Pengjun Zhou, Jingren Jiang, Yong Computation and Language Artificial Intelligence Confidence in LLMs is a useful indicator of model uncertainty and answer reliability. Existing work mainly focused on single-turn scenarios, while research on confidence in complex multi-turn interactions is limited. In this paper, we investigate whether LLM-based search agents have the ability to communicate their own confidence through verbalized confidence scores after long sequences of actions, a significantly more challenging task compared to outputting confidence in a single interaction. Experimenting on open-source agentic models, we first find that models exhibit much higher task accuracy at high confidence while having near-zero accuracy when confidence is low. Based on this observation, we propose Test-Time Scaling (TTS) methods that use confidence scores to determine answer quality, encourage the model to try again until reaching a satisfactory confidence level. Results show that our proposed methods significantly reduce token consumption while demonstrating competitive performance compared to baseline fixed budget TTS methods. |
| title | BrowseConf: Confidence-Guided Test-Time Scaling for Web Agents |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2510.23458 |