发表机构
University of Toronto; Microsoft Research（多伦多大学; 微软研究院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
研究多轮对话中语言模型响应的归因问题，提出多轮上下文归因及 Tokengeist 框架，通过有向无环图递归遍历恢复依赖路径，发布基准 MTCABench，实验表明 Tokengeist 能有效解决现有方法多跳依赖恢复不足问题。
AI 中文摘要
当语言模型在多轮对话中生成响应时，之前轮次的哪些 tokens 塑造了该答案，以及这些依赖关系如何在之前轮次中传播？现有上下文归因方法单次处理完整上下文，恢复表面级依赖但忽略现实对话和多步推理任务的分层、非线性结构。我们引入多轮上下文归因（MTCA），提出 Tokengeist，一个与归因方法无关且可扩展的框架，通过将归因转换为对对话轮次的有向无环图（DAG）的递归遍历恢复完整依赖路径。我们将发布 MTCABench，一个包含 665 个多轮对话中 3845 个目标跨度的基准，标注了深度达 14 的四种依赖类型的黄金溯源图。在四个开放权重模型上，扁平归因方法无法恢复多跳依赖，源召回率低于 20%，而 Tokengeist 达到 90%。我们的结果揭示了单次归因的系统失败模式——我们称之为溯源崩溃，并推动了跨轮次递归推理的归因方法。
英文摘要
When a language model produces a response in a multi-turn conversation, which tokens from prior turns shaped that answer, and how did those dependencies propagate across prior turns? Existing context attribution methods process the full context in a single pass, recovering surface-level dependencies but missing the layered, non-linear structure of real-world dialogues and multi-step reasoning tasks. We introduce multi-turn context attribution (MTCA): given a target span in a model response, the task of tracing attribution backward across turns to identify not only which prior turns were directly relevant, but also how those turns themselves depended on earlier context. We propose Tokengeist, an attribution-method-agnostic and scalable framework that recovers full dependency paths by casting attribution as a recursive traversal of a directed acyclic graph (DAG) over conversation turns. We will release MTCABench, a benchmark of 3,845 target spans across 665 multi-turn conversations, annotated with gold provenance graphs reaching depths of up to 14, across four dependency types. Across four open-weight models, flat attribution methods fail to recover multi-hop dependencies, achieving under 20% source recall, while Tokengeist reaches 90%. Our results reveal systematic failure modes of single-pass attribution -- which we term provenance collapse -- and motivate attribution methods that reason recursively across turns.
