通过集成多种方法提升大模型电路定位精度
BlackboxNLP-2025 MIB Shared Task: Exploring Ensemble Strategies for Circuit Localization Methods
- 并行与串行集成不同定位方法，融合边缘贡献度得分
- 串行集成用EAP-IG预热，再用剪枝法精细定位，提升精度
- 多方法并行集成效果最佳，适合追求高精度的可解释性研究
大规模语言模型（LLMs）中的电路定位任务旨在识别负责特定任务行为的子网络。本文探讨通过集成两种或更多电路定位方法是否能提升性能，提出并行与串行两种集成策略。并行集成通过平均、取最小值或最大值等方式融合不同方法对边的归因分数；串行集成则以EAP-IG获得的归因分数作为起点，驱动更耗时但更精确的边剪枝方法。实验表明，两种策略均显著提升基准指标表现，最终采用多方法并行集成（包括串行集成）取得最优结果。我们在BlackboxNLP 2025 MIB共享任务中评估该方法，在多个模型-任务组合下对比集成得分与官方基线。
原文摘要 · Abstract (English)
The Circuit Localization track of the Mechanistic Interpretability Benchmark (MIB) evaluates methods for localizing circuits within large language models (LLMs), i.e., subnetworks responsible for specific task behaviors. In this work, we investigate whether ensembling two or more circuit localization methods can improve performance. We explore two variants: parallel and sequential ensembling. In parallel ensembling, we combine attribution scores assigned to each edge by different methods-e.g., by averaging or taking the minimum or maximum value. In the sequential ensemble, we use edge attribution scores obtained via EAP-IG as a warm start for a more expensive but more precise circuit identification method, namely edge pruning. We observe that both approaches yield notable gains on the benchmark metrics, leading to a more precise circuit identification approach. Finally, we find that taking a parallel ensemble over various methods, including the sequential ensemble, achieves the best results. We evaluate our approach in the BlackboxNLP 2025 MIB Shared Task, comparing ensemble scores to official baselines across multiple model-task combinations.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。