---
title: "Attributing Agent Failures to Their History Needs an Executed Counterfactual: Toward a Reporting Standard"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2611.05650v1"
permalink: "https://feed.astralor.com/items/itm_a6855886d088cb7a32298433"
published: "2026-11-03T21:33:48.000Z"
score: 58
channel: "ai"
tags: ["AI Agent","Attributing Agent Failures","Their History Needs","Executed Counterfactual","Toward","Reporting Standard Published","Published","Each"]
---

# Attributing Agent Failures to Their History Needs an Executed Counterfactual: Toward a Reporting Standard

## At a glance

Published explanations of why long-horizon agents fail are commonly taxonomies labeled over completed trajectories: a judge reads a transcript and names the failure. We argue that a claim attributing an agent's behavior…

## Key points

- Published explanations of why long-horizon agents fail are commonly taxonomies labeled over completed trajectories: a judge reads a transcript and names the failure.
- We argue that a claim attributing an agent's behavior to its history should carry intervention-verified evidence, per instance or via a validated estimator: freeze an executable mid-trajectory state, alter only the hist…
- Each of the four manipulation classes we name has a precedent, and several methods already execute interventions from checkpoints, re-executed prefixes or recorded conversation states (for example DoVer, REFLECT, Min et…

## Full text

Published explanations of why long-horizon agents fail are commonly taxonomies labeled over completed trajectories: a judge reads a transcript and names the failure. We argue that a claim attributing an agent's behavior to its history should carry intervention-verified evidence, per instance or via a validated estimator: freeze an executable mid-trajectory state, alter only the history the agent sees, and continue - executed rather than inferred from the transcript, at a state whose restoration has been measured, with a replay certificate matched to the estimand. Each of the four manipulation classes we name has a precedent, and several methods already execute interventions from checkpoints, re-executed prefixes or recorded conversation states (for example DoVer, REFLECT, Min et al., Who&When Pro, Causal Agent Replay), but none reports a measured replay certificate. Two cautionary measurements, neither applying the standard, follow: a carrier taxonomy whose "concentration" could not be separated from benchmark construction (adjusted odds ratio 1.15, interval crossing 1), and hash-identical snapshots that, restored into identically seeded processes, agreed with one another in 24 of 24 states while differing from the live continuation in 9 and 20 of 24. Restoring the captured position of a random stream made all 24 agree with the live run.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_a6855886d088cb7a32298433