Temas / AI evaluation and generalizationPunto de vista parafraseado
Evolved harnesses generalize beyond their original benchmarks
A harness evolved on Terminal-Bench-2 transferred without further evolution to SWE-bench-verified, indicating it encodes general engineering experience in harness components rather than benchmark-specific optimization.
Detrás de la opinión
Las traducciones son para facilitar la lectura; los extractos originales siguen siendo la source evidence.
Harness Engineering for Self-Improvement
Extracto original
On Terminal-Bench-2, AHE achieved better than human-designed harness (OpenCode, Terminus-2, Codex) except for Hard tier and a few other self-evolve baselines (ACE, TF-GRPO). The same frozen harness, without further evolving, transfers to SWE-bench-verified, indicating that the evolved harness is able to encode engineering experience into harness components rather than doing benchmark-specific optimization.