Adapting While Learning: Grounding LLMs for Scientific Problems with Tool Usage Adaptation

Bohan Lyu, Yadi Cao, Duncan Watson-Parris, Leon Bergen, Taylor Berg-Kirkpatrick, Rose Yu
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:41701-41737, 2025.

Abstract

Large Language Models (LLMs) demonstrate promising capabilities in solving scientific problems but often suffer from the issue of hallucination. While integrating LLMs with tools can mitigate this issue, models fine-tuned on tool usage become overreliant on them and incur unnecessary costs. Inspired by how human experts assess problem complexity before selecting solutions, we propose a novel two-component fine-tuning method, Adapting while Learning (AWL). In the first component World Knowledge Learning (WKL), LLMs internalize scientific knowledge by learning from tool-generated solutions. In the second component Tool Usage Adaptation (TUA), we categorize problems as easy or hard based on the model’s accuracy, and train it to maintain direct reasoning for easy problems while switching to tools for hard ones. We validate our method on 6 scientific benchmark datasets across climate science, epidemiology, physics, and other domains. Compared to the original instruct model (8B), models post-trained with AWL achieve 29.11% higher answer accuracy and 12.72% better tool usage accuracy, even surpassing state-of-the-art models including GPT-4o and Claude-3.5 on 4 custom-created datasets. Our code is open-source at https://github.com/Rose-STL-Lab/Adapting-While-Learning.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-lyu25a, title = {Adapting While Learning: Grounding {LLM}s for Scientific Problems with Tool Usage Adaptation}, author = {Lyu, Bohan and Cao, Yadi and Watson-Parris, Duncan and Bergen, Leon and Berg-Kirkpatrick, Taylor and Yu, Rose}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {41701--41737}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/lyu25a/lyu25a.pdf}, url = {https://proceedings.mlr.press/v267/lyu25a.html}, abstract = {Large Language Models (LLMs) demonstrate promising capabilities in solving scientific problems but often suffer from the issue of hallucination. While integrating LLMs with tools can mitigate this issue, models fine-tuned on tool usage become overreliant on them and incur unnecessary costs. Inspired by how human experts assess problem complexity before selecting solutions, we propose a novel two-component fine-tuning method, Adapting while Learning (AWL). In the first component World Knowledge Learning (WKL), LLMs internalize scientific knowledge by learning from tool-generated solutions. In the second component Tool Usage Adaptation (TUA), we categorize problems as easy or hard based on the model’s accuracy, and train it to maintain direct reasoning for easy problems while switching to tools for hard ones. We validate our method on 6 scientific benchmark datasets across climate science, epidemiology, physics, and other domains. Compared to the original instruct model (8B), models post-trained with AWL achieve 29.11% higher answer accuracy and 12.72% better tool usage accuracy, even surpassing state-of-the-art models including GPT-4o and Claude-3.5 on 4 custom-created datasets. Our code is open-source at https://github.com/Rose-STL-Lab/Adapting-While-Learning.} }
Endnote
%0 Conference Paper %T Adapting While Learning: Grounding LLMs for Scientific Problems with Tool Usage Adaptation %A Bohan Lyu %A Yadi Cao %A Duncan Watson-Parris %A Leon Bergen %A Taylor Berg-Kirkpatrick %A Rose Yu %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-lyu25a %I PMLR %P 41701--41737 %U https://proceedings.mlr.press/v267/lyu25a.html %V 267 %X Large Language Models (LLMs) demonstrate promising capabilities in solving scientific problems but often suffer from the issue of hallucination. While integrating LLMs with tools can mitigate this issue, models fine-tuned on tool usage become overreliant on them and incur unnecessary costs. Inspired by how human experts assess problem complexity before selecting solutions, we propose a novel two-component fine-tuning method, Adapting while Learning (AWL). In the first component World Knowledge Learning (WKL), LLMs internalize scientific knowledge by learning from tool-generated solutions. In the second component Tool Usage Adaptation (TUA), we categorize problems as easy or hard based on the model’s accuracy, and train it to maintain direct reasoning for easy problems while switching to tools for hard ones. We validate our method on 6 scientific benchmark datasets across climate science, epidemiology, physics, and other domains. Compared to the original instruct model (8B), models post-trained with AWL achieve 29.11% higher answer accuracy and 12.72% better tool usage accuracy, even surpassing state-of-the-art models including GPT-4o and Claude-3.5 on 4 custom-created datasets. Our code is open-source at https://github.com/Rose-STL-Lab/Adapting-While-Learning.
APA
Lyu, B., Cao, Y., Watson-Parris, D., Bergen, L., Berg-Kirkpatrick, T. & Yu, R.. (2025). Adapting While Learning: Grounding LLMs for Scientific Problems with Tool Usage Adaptation. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:41701-41737 Available from https://proceedings.mlr.press/v267/lyu25a.html.

Related Material