Bad Genius: Counterfactual-Guided Harness Evolution Beyond Task-Specific Shortcuts

2026-09-16

AI Summary

论文提出Counterfactual Harness Search and Evolution方法，解决自动优化过程中产生的任务特定捷径问题，确保在协议变化下评估代理的有效性。

Abstract

Reliable agent evaluation is complicated by automatic harness optimization, which repeatedly uses a released benchmark $B_{\mathrm{rel}}$ to guide a Proposer that edits prompts, memory, retrieval, tools, and control code around a fixed target agent. Task holdout varies semantic tasks but leaves the benchmark protocol fixed, so a "bad genius" Proposer can produce a cheating harness whose released-benchmark gain depends on a benchmark-wide shortcut. We introduce Counterfactual Harness Search and Evolution (CHASE), which casts harness evolution as constraint generation over validity-preserving benchmark counterfactuals. After each Proposer update, a Challenger searches for an executable protocol transformation with large gain destruction. A validity firewall checks that task semantics are preserved, while a confirmation set determines whether the counterfactual enters a finite archive. We formalize an exact shortcut-neutralized benchmark $B_0$ and establish statistical guarantees linking finite counterfactual archives to $B_0$ and characterizing sequential Challenger search. We evaluate CHASE on a synthetic benchmark and on OfficeQA, where CHASE retains strong released-benchmark gains while substantially reducing gain destruction under valid protocol changes.

Problem

• Benchmark
• Harness Optimization
• Task-specific Shortcuts
• Agent Evaluation

Innovation

• Counterfactual Harness Search and Evolution
• validity-preserving benchmark counterfactuals
• shortcut-neutralized benchmark