AI4AI at Test-Time: Strong-to-Weak Capability Transfer via Harnesses
2026-08-12 • Machine Learning
Machine LearningArtificial IntelligenceComputation and Language
AI summaryⓘ
The authors explore whether a strong AI model can help a weaker one perform better during testing without changing the weaker model’s parameters. They create a system where the strong model builds a ‘harness’—a kind of framework or guide—that assists the weaker model in solving tasks more reliably. Testing on Theory-of-Mind tasks shows this approach almost doubles the weaker model's accuracy, mainly by turning uncertain parts of reasoning into fixed, rule-based steps. The results suggest that this test-time collaboration complements traditional training methods and especially benefits weaker models.
model distillationtest-time adaptationTheory of Mindinference harnessparameter updatesdeterministic codemodel reasoningtraining-time methodsAI collaborationcognitive structure transfer
Authors
Cheng Qian, Wenting Zhao, Liangwei Yang, Heng Wang, Jielin Qiu, Heng Ji, Silvio Savarese, Huan Wang, Shelby Heinecke
Abstract
Recent work on distillation transfers the capabilities of large models to smaller ones often by updating the latter's parameters, through teacher forcing, on-policy distillation, and related training-time methods. In this paper, we ask whether such transfer can instead occur at test time. We study strong-to-weak scaffolding: whether a stronger builder model can construct inference-time harnesses that help a weaker target model solve tasks more reliably without any parameter updates. Using four representative Theory-of-Mind benchmarks, each builder model uses 5% of the data as a validation set to iteratively refine its harness over multiple rounds, after which the finalized harness is evaluated on the full test set. Empirically, this form of test-time capability transfer is highly effective, nearly doubling average target-model performance from 0.49 to 0.91. Our analysis shows that the gains come primarily from offloading unstable model reasoning into deterministic code, benchmark-specific routing, and strict answer-format enforcement, rather than from encouraging the target model to reason more extensively or sample more broadly. We further find that builder-model reasoning effort improves harness quality monotonically, platform effects are modest relative to the builder model's own capability, and weaker target models receive the largest gains. These results suggest that inference-time harness design is an important complement to conventional training-time distillation, enabling strong models to transfer cognitive structure to weaker models without retraining.