CoTrace: Data Recipes for Training Terminal Agents with Harness-Model Co-Evolution

Key points

• Terminal-agent capability depends jointly on model weights and the runtime harness that formats prompts, binds tools, and handles error recovery.1
• Existing harness-model co-evolution approaches improve both components, yet often treat trajectories produced during harness search as an undifferentiated replay buffer.1
• To systematically analyze this interface, we establish an alternating co-evolution framework that decouples harness search and policy training through component-wise promotion decisions.1
• Under CoTrace, recurring execution failures guide harness synthesis, while policy training is strictly conditioned on verified rollouts matched to the adopted runtime for supervised fine-tuning (SFT) or fresh online interactions for reinforcement learning (RL).1

Sources (1)

• [1]CoTrace: Data Recipes for Training Terminal Agents with Harness-Model Co-Evolution
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 05:06 PM

  Within this framework, we introduce CoTrace, a harness-aware data recipe that explicitly governs trajectory routing, provenance matching, and curriculum refresh.

  Terminal-agent capability depends jointly on model weights and the runtime harness that formats prompts, binds tools, and handles error recovery.

Extractive summary: sentences quoted from the sources.