A self-improving coding-agent harness on pi: a Terminal-Bench agent whose harness learns from its own failures, plus Harness Learning (arXiv:2609.35738)
-
Updated
Oct 17, 2026 - TypeScript
#
harness-learning
Here are 3 public repositories matching this topic...
antar (अन्तर) — a self-improving multi-agent that finds the verified gap between your ML code and the literature, opens the issue, and codes the fix. Harness learning (arXiv:2609.35738), SOD-verified, fully traced. neatHack 2026.
python machine-learning code-analysis multi-agent arxiv observability rag ai-agent llm research-agent agentic-ai neatlogs harness-learning
-
Updated
Oct 12, 2026 - Python
Unofficial, credited reading companion to Harness Learning Enables Generalizable Test-Time Adaptation: method, protocols, results, caveats, and agent-ready source maps.
-
Updated
Oct 12, 2026 - Python
Add this topic to your repo
To associate your repository with the harness-learning topic, visit your repo's landing page and select "manage topics."