paper-with-me

홈 › Papers

AI4AI at Test-Time: Strong-to-Weak Capability Transfer via Harnesses

2026-08-12 · Cheng Qian, Wenting Zhao, Liangwei Yang, Heng Wang, Jielin Qiu, Heng Ji, Silvio Savarese, Huan Wang, Shelby Heinecke hf

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.

📄 PDF Abstract BibTeX arXiv:2608.12307

Code (2)

Tavish9/awesome-daily-AI-arxiv ★ 113
Valiant-Cat/hfpaper

Similar Papers 제목 키워드 기반

TTSR: Test-Time Self-Evolving via Reflection

2026-02-06 · Haoyang He, Zihua Rong, Liangjie Zhao, Yunjia Zhao 외 arxiv

Test-time training (TTT) adapts large language models (LLMs) during inference using only unlabeled test inputs. Existing methods, however, face two major bottlenecks on hard reasoning tasks: (1) \emph{lack of learnable s…

Mathematical Reasoning

Vanilla Transformers are Transfer Capability Teachers

2024-03-04 · Xin Lu, Yanyan Zhao, Bing Qin

Recently, Mixture of Experts (MoE) Transformers have garnered increasing attention due to their advantages in model capacity and computational efficiency. However, studies have indicated that MoE Transformers underperfor…

Computational EfficiencyMixture-of-Experts

When In-Distribution Gains Fail: Evaluating Weak-to-Strong Reward Models under Preference Shift

2026-05-25 · Khoi Le, Tri Cao, Phong Nguyen, Cong-Duy Nguyen 외 arxiv

Weak-to-strong (W2S) generalization is a promising framework for scalable oversight, yet existing evaluations often test students under matched train-test distributions. Therefore, we study W2S preference learning under …

Vision Superalignment: Weak-to-Strong Generalization for Vision Foundation Models

2024-02-06 · Jianyuan Guo, Hanting Chen, Chengcheng Wang, Kai Han 외

Recent advancements in large language models have sparked interest in their extraordinary and near-superhuman capabilities, leading researchers to explore methods for evaluating and optimizing these abilities, which is c…

Few-Shot LearningKnowledge DistillationTransfer Learning

Contrastive Weak-to-strong Generalization

2025-10-09 · Houcheng Jiang, Junfeng Fang, Jiaxin Wu, Tianyu Zhang 외 arxiv

Weak-to-strong generalization provides a promising paradigm for scaling large language models (LLMs) by training stronger models on samples from aligned weaker ones, without requiring human feedback or explicit reward mo…