新方法Delta-Crosscoder能精准定位微调引发的模型变化，尤其适合小范围微调场景。
Delta-Crosscoder: Robust Crosscoder Model Diffing in Narrow Fine-Tuning Regimes
- 用增量损失和稀疏化聚焦微调中真正变化的隐层方向
- 在10个模型上验证，能准确识别导致行为改变的因果方向
- 适合研究模型微调副作用或可解释性的研究人员
模型差分方法旨在识别微调如何改变模型内部表征。交叉编码器通过学习基础模型与微调后模型之间的可解释隐层方向共享字典来实现这一目标。然而，现有方法在窄范围微调场景下表现不佳，此时行为变化局部且不对称。本文提出Delta-Crosscoder，结合BatchTopK稀疏性、基于增量的损失以优先关注模型间变化的方向，并利用配对输入上的激活隐含对比信号。在包括合成假信息、新兴错位、潜意识学习及禁忌词猜测在内的10个模型（Gemma、LLaMA、Qwen；参数量1B-9B）上评估，该方法能可靠识别导致微调行为变化的因果隐层方向，并有效用于缓解问题，优于SAE基线，与非SAE基线相当。结果表明，交叉编码器仍是模型差分的强大工具。
原文摘要 · Abstract (English)
Model diffing methods aim to identify how fine-tuning changes a model's internal representations. Crosscoders approach this by learning shared dictionaries of interpretable latent directions between base and fine-tuned models. However, existing formulations struggle with narrow fine-tuning, where behavioral changes are localized and asymmetric. We introduce Delta-Crosscoder, which combines BatchTopK sparsity with a delta-based loss prioritizing directions that change between models, plus an implicit contrastive signal from paired activations on matched inputs. Evaluated across 10 model organisms, including synthetic false facts, emergent misalignment, subliminal learning, and taboo word guessing (Gemma, LLaMA, Qwen; 1B-9B parameters), Delta-Crosscoder reliably isolates latent directions causally responsible for fine-tuned behaviors and enables effective mitigation, outperforming SAE-based baselines, while matching the Non-SAE-based. Our results demonstrate that crosscoders remain a powerful tool for model diffing.
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。