arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2605.12809cs.LGcs.AI
收藏

修正影响:通过正交潜在空间解包LLM输出

Correcting Influence: Unboxing LLM Outputs with Orthogonal Latent Spaces

Electrical and Computer Engineering(电气与计算机工程系);Department of Statistics(统计学系);Operations Research and Industrial Engineering(运筹学与工业工程)

Shixing Yu, Promit Ghosal, Kyra Gan

AI总结:

本文提出一种通过潜在中介方法推断token级影响的框架,用于医疗任务中提升LLM预测的可解释性和可信度。

摘要:

可靠使用大语言模型(LLM)于医疗的关键步骤是将预测归因于训练数据,类似于医学病例研究。这需要逐token精度:不仅确定哪些训练示例影响决策,还要确定其中哪些token负责。虽然影响函数提供了一个原理性框架,但先前工作受限于自回归设置,并假设token独立性,导致识别的影响不可靠。我们引入了一个灵活的框架,通过潜在中介方法推断token级影响,适用于一般预测任务。我们的方法将稀疏自编码器附加到预训练LLM的任意层,以学习近似独立的潜在特征基础。不同于先前方法中token级影响加法分解,基于潜在特征的影响本质上非可分解。为此,我们引入了一种使用雅可比-向量积的新型方法。通过token激活模式将潜在归因传播回输入空间以获得token级影响。我们通过高效的逆Hessian近似扩展了我们的方法。在医疗基准上的实验表明,我们的方法识别出稀疏、可解释的token集,共同影响预测。我们的框架增强了信任并使模型审计成为可能,适用于需要透明和可问责决策的高风险领域。

英文摘要:

A critical step for reliable large language models (LLMs) use in healthcare is to attribute predictions to their training data, akin to a medical case study. This requires token-level precision: pinpointing not just which training examples influence a decision, but which tokens within them are responsible. While influence functions offer a principled framework for this, prior work is restricted to autoregressive settings and relies on an implicit assumption of token independence, rendering their identified influences unreliable. We introduce a flexible framework that infers token-level influence through a latent mediation approach for general prediction tasks. Our method attaches sparse autoencoders to any layer of a pretrained LLM to learn a basis of approximately independent latent features. Unlike prior methods where influence decomposes additively across tokens, influence computed over latent features is inherently non-decomposable. To address this, we introduce a novel method using Jacobian-vector products. Token-level influence is obtained by propagating latent attributions back to the input space via token activation patterns. We scale our approach using efficient inverse-Hessian approximations. Experiments on medical benchmarks show our approach identifies sparse, interpretable sets of tokens that jointly influence predictions. Our framework enhances trust and enables model auditing, generalizing to high-stakes domain requiring transparent and accountable decisions.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑