GH-AWTrace: 25,014 GitHub Agentic Workflows Linked to 4.1M Run Records
ETSMTL · hf · 2026-11-16
Researchers at ÉTS Montréal, Queen's University, and Laval University present GH-AWTrace, a large-scale dataset of GitHub Agentic Workflow execution logs, addressing the gap that prior studies examine workflow definitions and outputs but offer little evidence of what actually happens during execution.
- Scale: links 25,014 verified GitHub Agentic Workflows across 5,051 repositories to 4,112,526 run records and 562,877 nonempty execution log archives.
- Execution evidence is preserved within GitHub's log-retention limits and linked to repository metadata and generated workflow configurations.
- Illustrative cases show logs exposing tool denials, provider failures, shifts in an agent's approach, and delegation among agents.
- Enables research on execution failures, tool use, and the relationship between reported run outcomes and visible task progress. Coverage varies across workflows; logs are partial observations, not complete agent trajectories.
More from coding & agent
- Open-source Ontology Atlas gives humans and coding agents one shared Markdown codebase ontology — adnan_hashmi · 2026-11-16
- Engineer runs 4,000 agent PRs a month in free 1-hour GrokBot team masterclass — gaganghotra_ · 2026-11-16
- Specialized 4B models suffice for most tasks, and even 350M overperforms — Apprehensive_Ring666 · 2026-11-16
- Windows 11 build caught probing for Codex, Claude, Cursor agents via ETW telemetry — ssh4net · 2026-11-16
- Google-led T2J-Bench and TwinHarness lift PyTorch-to-JAX migration reward 11.4% across 63 repo tasks — jiqizhixin · 2026-11-16
- Designer ships Site Tuner: Figma-style editor for fine-tuning agent-built pages — jashangupta2009 · 2026-11-16