HuggingFace Daily Papers·· 2 天前
论文提出 CREDIT：用反事实训练提升语言模型自我归因能力
Did the Model Use That? Counterfactual Training Boosts Self-Attribution in Language Models
AI 导读
论文提出 CREDIT 方法，通过反事实标签直接监督模型最终答案的自我归因，无需人工标注。研究发现 11 个前沿模型中多数很少标注影响答案的外部信息，其余则标注了无影响的信息，且将“承认”训练进推理过程也无法传递到最终答案。CREDIT 在四个开源模型（含两个蒸馏推理模型）上显著提升归因表现，优于推理层微调与提示词基线。
正文 · 原文
Abstract:A fundamental challenge for explainable and accountable language models is revealing how external information influences their answers. This paper examines whether models are self-attributive: that is, whether they indicate when -- and only when -- their responses to questions are influenced by information provided alongside the question. Furthermore, we explore how self-attribution can be improved through counterfactual training, which explicitly teaches models to distinguish between answers that do and do not depend on supplied information.
We found that most of the eleven frontier models rarely attribute influential information in their answers, and the rest attribute non-influential information. Training acknowledgment into reasoning does not fix this: the acknowledgment does not reach the final answer.
We introduce CREDIT, which directly supervises final-answer self-attribution with counterfactual labels, without the need for human annotation. CREDIT substantially improves attribution in four open models, including two distilled reasoning models, and outperforms reasoning-level fine-tuning and prompting baselines. Among answers that agree with a user's suggestion, CREDIT distinguishes those changed by the suggestion from those the model would have given anyway. Gains transfer partly to unseen information types and prompts containing sources that support conflicting answers.
| Comments: | 29 pages, 5 figures, 20 tables. Code: this https URL |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| ACM classes: | I.2.7 |
| Cite as: | arXiv:2610.35295 [cs.CL] |
| (or arXiv:2610.35295v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.35295 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Zikai Zhou [view email]
[v1]
Mon, 26 Oct 2026 12:03:05 UTC (222 KB)
来源：HuggingFace Daily Papers · arxiv.org