Reference implementation for Continual Harness (Karten et al., 2026), a reset-free framework that automates agentic-harness refinement through online in-context learning, evaluated on Pokémon Red and ...
2026-06-25 🧪 Strict validation JSON + calmer agent context: standalone backtest validation now normalizes nested NaN / Infinity values before writing artifacts/validation.json or CLI stdout, so ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results