Adapting LLM agents to domain-specific tool calling remains notably brittle under evolving interfaces. Prompt and schema engineering is easy to deploy but often fragile under distribution shift and strict parsers, while continual parameter-efficient fine-tuning improves reliability at the cost of training, maintenance, and potential forgetting. We identify a critical Lazy Agent failure mode where tool necessity is nearly perfectly decodable from mid-layer activations, yet the model remains conservative in entering tool mode, revealing a representation-behavior gap. We propose Activation Steering Adapter (ASA), a training-free, inference-time controller that performs a single-shot mid-layer intervention and targets tool domains via a router-conditioned mixture of steering vectors with a probe-guided signed gate to amplify true intent while suppressing spurious triggers. On MTU-Bench with Qwen2.5-1.5B, ASA improves strict tool-use F1 from 0.18 to 0.50 while reducing the false positive rate from 0.15 to 0.05, using only about 20KB of portable assets and no weight updates.
@article{arxiv.2602.04935,
title = {ASA: Training-Free Representation Engineering for Tool-Calling Agents},
author = {Youjin Wang and Run Zhou and Rong Fu and Shuaishuai Cao and Hongwei Zeng and Jiaxuan Lu and Sicheng Fan and Jiaqiao Zhao and Liangming Pan},
journal= {arXiv preprint arXiv:2602.04935},
year = {2026}
}