让大模型能追踪行为来源，快速定位问题数据。
DebugLM: Learning Traceable Training Data Provenance for LLMs
- 给模型注入唯一数据溯源标签，关联回答与训练数据源。
- 可精准识别不良行为来源，支持不重训的即时拒绝策略。
- 适合需要可解释性与可控性的大模型开发团队。
大型语言模型在多阶段流水线中使用异构数据源进行训练，但开发者缺乏有效手段定位导致特定行为的具体数据。这种可观察性的缺失使调试沦为被动修补，且在分布偏移或后续模型更新时故障易复发。为此，我们提出 DebugLM 框架，为 LLM 引入内置数据溯源能力，使其能明确追溯行为来源至特定训练数据集。具体而言，模型学习将响应与唯一溯源标签关联，标识出负责该行为的数据源，帮助开发者精确定位不良行为的学习来源。基于此能力，DebugLM 进一步支持测试时的定向修复，可在不重新训练或修改模型参数的前提下，对指定数据源触发选择性拒绝。实验表明，DebugLM 在多阶段训练流程中实现了准确的行为溯源，并有效支持测试时修复，同时保持模型通用能力。
原文摘要 · Abstract (English)
Large language models (LLMs) are trained through multi-stage pipelines over heterogeneous data sources, yet developers lack a principled way to pinpoint the specific data responsible for an observed behavior. This lack of observability reduces debugging to reactive patching and makes failures prone to recur under distribution shift or subsequent model updates. To address this limitation, we propose DebugLM, a framework that equips LLMs with built-in data provenance, enabling them to explicitly trace the origins of their behaviors to specific training data sources. Specifically, the model learns to associate its responses with unique provenance tags that indicate the responsible dataset, empowering developers to precisely identify where undesirable behaviors are learned. Building on this capability, DebugLM further supports targeted test-time remediation, enabling developers to selectively trigger targeted refusal for specified data sources without retraining or modifying model parameters. Experiments demonstrate that DebugLM provides accurate behavior tracing in multi-stage training pipelines and effective test-time remediation while preserving the general utility of the model.
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。