Search, Do not Guess: Teaching Small Language Models to Be Effective Search Agents
Yizhou Liu, Qi Sun, Yulin Chen, Siyue Zhang, Chen Zhao · 2026
Read Search, Do not Guess: Teaching Small Language Models to Be Effective Search Agents by Yizhou Liu, Qi Sun, Yulin Chen, Siyue Zhang, Chen Zhao (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-04651.txt.
Description
Liu and four colleagues teach small language models to search rather than guess. Large language models reason strongly but cost much. Yet on multi-hop reasoning tasks, small models call search tools less often and hallucinate more. The team proposes lightweight fine-tuning training small models to retrieve and generate answers grounded in evidence. Against agent distillation from large models, the approach gains 17.3 on Bamboogle and 15.3 on HotpotQA, reaching large-model-level results. Adaptive search strategies in small models degrade performance, showing consistent search behavior matters.
Catalog record
| CREATOR | Yizhou Liu, Qi Sun, Yulin Chen, Siyue Zhang, Chen Zhao |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 39237 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.04651