Researchers from multiple institutions introduced Meta-Harness, an outer-loop system that searches over harness code for LLM applications. Rather than relying on hand-designed harnesses or text optimizers that compress feedback too aggressively, Meta-Harness uses an agentic proposer that accesses the source code, scores, and execution traces of all prior candidates through a filesystem.
On online text classification, Meta-Harness improved over a state-of-the-art context management system by 7.7 points while using 4x fewer context tokens. On retrieval-augmented math reasoning, a single discovered harness improved accuracy on 200 IMO-level problems by 4.7 points on average across five held-out models. On agentic coding, discovered harnesses surpassed the best hand-engineered baselines on TerminalBench-2.
Harness design has been a largely manual process, with existing text optimizers poorly suited to the task. Meta-Harness demonstrates that richer access to prior experience can enable automated harness engineering, potentially accelerating LLM system development and reducing the engineering burden of context management and tool orchestration.
Sources: Source 1