Discovering Transformer Circuits via a Hybrid Attribution and Pruning Framework
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908574921261056 |
|---|---|
| author | Gu, Hao Nair, Vibhas Kumar, Amrithaa Ashok Sharma, Jayvart Lagasse, Ryan |
| author_facet | Gu, Hao Nair, Vibhas Kumar, Amrithaa Ashok Sharma, Jayvart Lagasse, Ryan |
| contents | Interpreting language models often involves circuit analysis, which aims to identify sparse subnetworks, or circuits, that accomplish specific tasks. Existing circuit discovery algorithms face a fundamental trade-off: attribution patching is fast but unfaithful to the full model, while edge pruning is faithful but computationally expensive. This research proposes a hybrid attribution and pruning (HAP) framework that uses attribution patching to identify a high-potential subgraph, then applies edge pruning to extract a faithful circuit from it. We show that HAP is 46\% faster than baseline algorithms without sacrificing circuit faithfulness. Furthermore, we present a case study on the Indirect Object Identification task, showing that our method preserves cooperative circuit components (e.g. S-inhibition heads) that attribution patching methods prune at high sparsity. Our results show that HAP could be an effective approach for improving the scalability of mechanistic interpretability research to larger models. Our code is available at https://anonymous.4open.science/r/HAP-circuit-discovery. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_03282 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Discovering Transformer Circuits via a Hybrid Attribution and Pruning Framework Gu, Hao Nair, Vibhas Kumar, Amrithaa Ashok Sharma, Jayvart Lagasse, Ryan Machine Learning Computation and Language I.2.6; I.2.7 Interpreting language models often involves circuit analysis, which aims to identify sparse subnetworks, or circuits, that accomplish specific tasks. Existing circuit discovery algorithms face a fundamental trade-off: attribution patching is fast but unfaithful to the full model, while edge pruning is faithful but computationally expensive. This research proposes a hybrid attribution and pruning (HAP) framework that uses attribution patching to identify a high-potential subgraph, then applies edge pruning to extract a faithful circuit from it. We show that HAP is 46\% faster than baseline algorithms without sacrificing circuit faithfulness. Furthermore, we present a case study on the Indirect Object Identification task, showing that our method preserves cooperative circuit components (e.g. S-inhibition heads) that attribution patching methods prune at high sparsity. Our results show that HAP could be an effective approach for improving the scalability of mechanistic interpretability research to larger models. Our code is available at https://anonymous.4open.science/r/HAP-circuit-discovery. |
| title | Discovering Transformer Circuits via a Hybrid Attribution and Pruning Framework |
| topic | Machine Learning Computation and Language I.2.6; I.2.7 |
| url | https://arxiv.org/abs/2510.03282 |