Thinker: Training LLMs in Hierarchical Thinking for Deep Search via Multi-Turn Interaction
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866918201957285888 |
|---|---|
| author | Xu, Jun Du, Xinkai Ao, Yu Zhao, Peilong Li, Yang Zhong, Ling Yuan, Lin Bo, Zhongpu Wang, Xiaorui Sun, Mengshu Gui, Zhengke Zhang, Dalong Wang, Zhaoyang Wang, Qiwei Hou, Yangyang Yin, Zhiying Wang, Haofen Chen, Huajun Liang, Lei Zhou, Jun |
| author_facet | Xu, Jun Du, Xinkai Ao, Yu Zhao, Peilong Li, Yang Zhong, Ling Yuan, Lin Bo, Zhongpu Wang, Xiaorui Sun, Mengshu Gui, Zhengke Zhang, Dalong Wang, Zhaoyang Wang, Qiwei Hou, Yangyang Yin, Zhiying Wang, Haofen Chen, Huajun Liang, Lei Zhou, Jun |
| contents | Efficient retrieval of external knowledge bases and web pages is crucial for enhancing the reasoning abilities of LLMs. Previous works on training LLMs to leverage external retrievers for solving complex problems have predominantly employed end-to-end reinforcement learning. However, these approaches neglect supervision over the reasoning process, making it difficult to guarantee logical coherence and rigor. To address these limitations, we propose Thinker, a hierarchical thinking model for deep search through multi-turn interaction, making the reasoning process supervisable and verifiable. It decomposes complex problems into independently solvable sub-problems, each dually represented in both natural language and an equivalent logical function to support knowledge base and web searches. Concurrently, dependencies between sub-problems are passed as parameters via these logical functions, enhancing the logical coherence of the problem-solving process. To avoid unnecessary external searches, we perform knowledge boundary determination to check if a sub-problem is within the LLM's intrinsic knowledge, allowing it to answer directly. Experimental results indicate that with as few as several hundred training samples, the performance of Thinker is competitive with established baselines. Furthermore, when scaled to the full training set, Thinker significantly outperforms these methods across various datasets and model sizes. The source code is available at https://github.com/OpenSPG/KAG-Thinker. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_07943 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Thinker: Training LLMs in Hierarchical Thinking for Deep Search via Multi-Turn Interaction Xu, Jun Du, Xinkai Ao, Yu Zhao, Peilong Li, Yang Zhong, Ling Yuan, Lin Bo, Zhongpu Wang, Xiaorui Sun, Mengshu Gui, Zhengke Zhang, Dalong Wang, Zhaoyang Wang, Qiwei Hou, Yangyang Yin, Zhiying Wang, Haofen Chen, Huajun Liang, Lei Zhou, Jun Artificial Intelligence Computation and Language Efficient retrieval of external knowledge bases and web pages is crucial for enhancing the reasoning abilities of LLMs. Previous works on training LLMs to leverage external retrievers for solving complex problems have predominantly employed end-to-end reinforcement learning. However, these approaches neglect supervision over the reasoning process, making it difficult to guarantee logical coherence and rigor. To address these limitations, we propose Thinker, a hierarchical thinking model for deep search through multi-turn interaction, making the reasoning process supervisable and verifiable. It decomposes complex problems into independently solvable sub-problems, each dually represented in both natural language and an equivalent logical function to support knowledge base and web searches. Concurrently, dependencies between sub-problems are passed as parameters via these logical functions, enhancing the logical coherence of the problem-solving process. To avoid unnecessary external searches, we perform knowledge boundary determination to check if a sub-problem is within the LLM's intrinsic knowledge, allowing it to answer directly. Experimental results indicate that with as few as several hundred training samples, the performance of Thinker is competitive with established baselines. Furthermore, when scaled to the full training set, Thinker significantly outperforms these methods across various datasets and model sizes. The source code is available at https://github.com/OpenSPG/KAG-Thinker. |
| title | Thinker: Training LLMs in Hierarchical Thinking for Deep Search via Multi-Turn Interaction |
| topic | Artificial Intelligence Computation and Language |
| url | https://arxiv.org/abs/2511.07943 |