Deliberative Searcher: Improving LLM Reliability via Reinforcement Learning with constraints
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866915941086920704 |
|---|---|
| author | Yin, Zhenyun Wang, Shujie Wang, Xuhong Ma, Xingjun Wang, Yinchun |
| author_facet | Yin, Zhenyun Wang, Shujie Wang, Xuhong Ma, Xingjun Wang, Yinchun |
| contents | Improving the reliability of large language models (LLMs) is critical for deploying them in real-world scenarios. In this paper, we propose \textbf{Deliberative Searcher}, the first framework to integrate certainty calibration with retrieval-based search for open-domain question answering. The agent performs multi-step reflection and verification over Wikipedia data and is trained with a reinforcement learning algorithm that optimizes for accuracy under a soft reliability constraint. Empirical results show that proposed method improves alignment between model confidence and correctness, leading to more trustworthy outputs. This paper will be continuously updated. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2507_16727 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Deliberative Searcher: Improving LLM Reliability via Reinforcement Learning with constraints Yin, Zhenyun Wang, Shujie Wang, Xuhong Ma, Xingjun Wang, Yinchun Artificial Intelligence Improving the reliability of large language models (LLMs) is critical for deploying them in real-world scenarios. In this paper, we propose \textbf{Deliberative Searcher}, the first framework to integrate certainty calibration with retrieval-based search for open-domain question answering. The agent performs multi-step reflection and verification over Wikipedia data and is trained with a reinforcement learning algorithm that optimizes for accuracy under a soft reliability constraint. Empirical results show that proposed method improves alignment between model confidence and correctness, leading to more trustworthy outputs. This paper will be continuously updated. |
| title | Deliberative Searcher: Improving LLM Reliability via Reinforcement Learning with constraints |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2507.16727 |