Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2603.18474 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914460813230080 |
|---|---|
| author | Yu, Haonan Liu, Junhao Yan, Zhenyu Lin, Haoran Zhang, Xin |
| author_facet | Yu, Haonan Liu, Junhao Yan, Zhenyu Lin, Haoran Zhang, Xin |
| contents | Precise behavioral control of large language models (LLMs) is critical for complex applications. However, existing methods often incur high training costs, lack natural language controllability, or compromise semantic coherence. To bridge this gap, we propose WASD (unWeaving Actionable Sufficient Directives), a novel framework that explains model behavior by identifying sufficient neural conditions for token generation. Our method represents candidate conditions as neuron-activation predicates and iteratively searches for a minimal set that guarantees the current output under input perturbations. Experiments on SST-2 and CounterFact with the Gemma-2-2B model demonstrate that our approach produces explanations that are more stable, accurate, and concise than conventional attribution graphs. Moreover, through a case study on controlling cross-lingual output generation, we validated the practical effectiveness of WASD in controlling model behavior. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_18474 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | WASD: Locating Critical Neurons as Sufficient Conditions for Explaining and Controlling LLM Behavior Yu, Haonan Liu, Junhao Yan, Zhenyu Lin, Haoran Zhang, Xin Computation and Language Artificial Intelligence Precise behavioral control of large language models (LLMs) is critical for complex applications. However, existing methods often incur high training costs, lack natural language controllability, or compromise semantic coherence. To bridge this gap, we propose WASD (unWeaving Actionable Sufficient Directives), a novel framework that explains model behavior by identifying sufficient neural conditions for token generation. Our method represents candidate conditions as neuron-activation predicates and iteratively searches for a minimal set that guarantees the current output under input perturbations. Experiments on SST-2 and CounterFact with the Gemma-2-2B model demonstrate that our approach produces explanations that are more stable, accurate, and concise than conventional attribution graphs. Moreover, through a case study on controlling cross-lingual output generation, we validated the practical effectiveness of WASD in controlling model behavior. |
| title | WASD: Locating Critical Neurons as Sufficient Conditions for Explaining and Controlling LLM Behavior |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2603.18474 |