Tool Calling Interpretability

ASA

arxiv.org

https://arxiv.org/pdf/2602.04935v2

tool hallucination

arxiv.org

https://arxiv.org/pdf/2601.05214

trigger

aclanthology.org

https://aclanthology.org/2025.acl-long.655.pdf

RAG

CtrlA: Adaptive Retrieval-Augmented Generation via Inherent Control

Liu Huanshuo, Hao Zhang, Zhijiang Guo, Jing Wang, Kuicai Dong, Xiangyang Li, Yi Quan Lee, Cong Zhang, Yong Liu. Findings of the Association for Computational Linguistics: ACL 2025. 2025.

https://aclanthology.org/2025.findings-acl.652/

Probing-RAG: Self-Probing to Guide Language Models in Selective Document Retrieval

Ingeol Baek, Hwan Chang, ByeongJeong Kim, Jimin Lee, Hwanhee Lee. Findings of the Association for Computational Linguistics: NAACL 2025. 2025.

https://aclanthology.org/2025.findings-naacl.181/

Where prior work largely focused on detecting tool-calling errors or on binary choice control, this paper is the first to identify a linear switch toward a specific tool in multi-tool settings, along with the underlying mechanism: a three-stage circuit.

Tool Calling is Linearly Readable and Steerable in Language Models

When a tool-calling agent picks the wrong tool, the failure is invisible until execution: the email gets sent, the meeting gets missed. As agents take on consequential actions, one bad tool call...

https://arxiv.org/abs/2605.07990

Tool Calling Interpretability

ASA

Recommendations