WRSWorld Research StateContribute
WORLD RESEARCH STATE

Mistral AI and All Hands AI: Devstral coding-agent trajectory training

Contribution grade

Software-agent training systems · Human-designed research method or resource

Reference score55,000Provisional editorial estimate of this exact contribution; its grade is assessed separately.

What changed

Specialized an open language model for repository-level software agents using test-checked environment trajectories, two-stage quality filtering and further policy optimization within an established coding-agent scaffold.

Scientific scope

Specialized an open language model for repository-level software agents using test-checked environment trajectories, two-stage quality filtering and further policy optimization within an established coding-agent scaffold. OpenHands/CodeAct, SWE-Gym, vLLM and existing model architecture are inherited. The integrated training/resource is acquired, not invention of autonomous coding agents. Evaluation depends on scaffold, retry protocol, budgets and tests. Credited lists repeat one name and contain merged fields; individual identity/effort is unresolved.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. OpenHands/CodeAct, SWE-Gym, vLLM and existing model architecture are inherited. The integrated training/resource is acquired, not invention of autonomous coding agents. Evaluation depends on scaffold, retry protocol, budgets and tests. Credited lists repeat one name and contain merged fields; individual identity/effort is unresolved.

Contributors

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction