From Agent Traces to Trust: A Survey of Evidence Tracing and Execution Provenance in LLM Agents

Summary

Argues final-answer accuracy alone cannot explain how an output was produced, which evidence supported each claim, whether tool calls were justified, or where failures originated, and formalises execution provenance as the typed graph of an agent execution with evidence tracing as its projection onto evidence-support relations. Supplies a taxonomy of trace sources, evidence and execution units, provenance relations, granularity and timing, representation forms and trust functions, plus benchmarks, datasets and metrics. The reference for the paper's transparency/traceability outcome dimension, giving it an existing formal vocabulary rather than an ad hoc measure.

Contribution

Queued — not yet read.

Method

Queued — not yet read.

Relevance to RISE

Surfaced by the 2026-08-29 literature sweep as a medium-priority candidate for the Architecture as Epistemology (ISR) paper (angle: epistemic quality of LLM outputs). Verified against: Opened https://arxiv.org/abs/2606.04990 (arXiv abstract page, v4 28 Jun 2026)

Critique / open questions

Queued — not yet assessed.

Key quotes

Large language model (LLM)-based agents are evolving from passive text generators into autonomous systems capable of planning, tool use, retrieval, memory access, environmental interaction, and multi-agent collaboration. (abstract)

These capabilities expand agent autonomy, but also make agent behavior harder to verify, debug, and audit. (abstract)