research
The Rerun Reproduced Failure More Often Than It Repaired It
SymTrace replays multi-agent trajectories from intervention anchors to separate causal repair from lucky resampling.
Summary
SymTrace replays multi-agent trajectories from intervention anchors to separate causal repair from lucky resampling.
Across 536 human-annotated failures in three frameworks, unguided reruns reproduced failures 67.97 percent of the time but repaired only 6.90 percent. A symptom-driven intervention repaired 20.15 percent, a reported 191.89 percent improvement over the studied repair methods while still leaving most failures unresolved.
Why it matters
SymTrace replays multi-agent trajectories from intervention anchors to separate causal repair from lucky resampling.
Limits and context
- Across 536 human-annotated failures in three frameworks, unguided reruns reproduced failures 67.97 percent of the time but repaired only 6.90 percent.
Key claims
SymTrace replays multi-agent trajectories from intervention anchors to separate causal repair from lucky resampling.
Qualification: Across 536 human-annotated failures in three frameworks, unguided reruns reproduced failures 67.97 percent of the time but repaired only 6.90 percent.
Evidence: source-2026-08-27-007
Sources
- arXiv preprint 2608.25920arXiv · primary research
Corrections
No corrections have been recorded for this story.