跳到正文
arXiv cs.CL· Jessica Tang, Shraddha Barke, Sharad Agarwal·· 10 小时前AI 评分33

Tokengeist:智能体对话中的多轮归因追踪框架

Tokengeist: Multi-Turn Attribution Tracing in Agentic Conversations

AI 导读

Tokengeist 是一个与归因方法无关的可扩展框架,将归因建模为对话轮次有向无环图的递归遍历,以恢复完整依赖路径。在四个开放权重模型上,源召回率从扁平归因的 11-38% 提升至 82-94%。研究同时提出 MTCABench 基准,含 685 段多轮对话中的 3,970 个目标片段,标注深度最高达 17 的溯源图,并揭示了单遍归因的“溯源坍缩”失效模式。

正文 · 原文

View PDF HTML (experimental)

Abstract:When a model answers after several turns of dialogue and tool use, an overseer needs to know which earlier turns and intermediate steps it rests on. Attribution is the natural primitive for answering this, but existing context attribution methods process the full transcript in a single pass: they recover surface-level dependencies while missing the layered structure of multi-turn traces, in which an assistant turn is itself a derived artifact that condenses or transforms earlier context. 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 over conversation turns. We will release MTCABench, a benchmark of 3,970 target spans across 685 multi-turn conversations, annotated with gold provenance graphs reaching depths of up to 17, across four dependency types. Across four open-weight models, flat attribution methods fail to recover multi-hop dependencies, with source recall improving from 11-38% under flat attribution to 82-94% under Tokengeist. Our results reveal a systematic failure mode of single-pass attribution, which we term provenance collapse, where attribution terminates at the nearest assistant restatement instead of the originating event.
Subjects: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
Cite as: arXiv:2607.22610 [cs.AI]
  (or arXiv:2607.22610v2 [cs.AI] for this version)
  https://doi.org/10.48550/arXiv.2607.22610

arXiv-issued DOI via DataCite

Submission history

From: Jessica Tang [view email]
[v1] Sun, 14 Jun 2026 17:14:01 UTC (1,452 KB)
[v2] Thu, 19 Nov 2026 16:06:35 UTC (1,635 KB)

来源:arXiv cs.CL · arxiv.org