Test-time training (TTT) adapts language models through gradient-based updates at inference. But is adaptation the right strategy? We study compute-optimal test-time strategies for verifiable execution-grounded (VEG) tasks, domains like GPU kernel optimization where a deterministic evaluator provides dense, continuous reward signals. Using KernelBench as our testbed and a 120B-parameter model (GPT-OSS-120B with LoRA adaptation), we find that search outperforms minimal adaptation (1-5 gradient steps): Best-of-N sampling achieves 90% task success (18/20 tasks) at K=64 across the full KernelBench L1 eval set while TTT's best checkpoint reaches only 30.6% (3-seed mean), with TTT's "equivalent K" falling below 1, worse than single-sample inference. The failure mode is over-sharpening: gradient updates collapse diversity toward mediocre solutions rather than discovering optimal ones. Our main contribution is surprisal-guided selection: selecting the highest-surprisal (lowest-confidence) correct sample yields 80% success vs. 50% for most-confident selection, a 30% improvement. Extending to surprisal-guided-top3 matches oracle performance at 100%. This zero-cost strategy, validated through length-controlled analysis, recovers oracle performance. For dense-reward VEG tasks, compute should be allocated to sample diversity and intelligent selection rather than gradient adaptation. The surprisal-guided selection principle may generalize to other execution-grounded domains where optimal solutions occupy the distribution tail.
FactReview: Evidence-Grounded Peer Review with Execution-Based Claim Verification
Rensselaer Polytechnic Institute（罗切斯特理工学院）；Sun Yat-sen University（中山大学）；Southeast University（东南大学）；The Hong Kong University of Science and Technology（香港科技大学）
MOSAIC: Modular Orchestration for Structured Agentic Intelligence and Composition
Department of Computer Science, National University of Singapore（新加坡国立大学计算机科学系）；University College London（伦敦大学学院）；University of Edinburgh（爱丁堡大学）；Data & Analytics, Digital X（Digital X 数据与分析部）；Alan Turing Institute（艾伦·图灵研究所）
MARBLE: Multi-Agent Reasoning for Bioinformatics Learning and Evolution
Division of AI Convergence, Dongguk University, Seoul, South Korea（东国大学人工智能融合系）；AI Research Team, Ar-ge Inc., Seoul, South Korea（Ar-ge公司人工智能研究团队）；Department of Biomedical Engineering, Dongguk University, Goyang-si, Gyeonggi-do, South Korea（东国大学生物医学工程系）；Department of Computer Science and Artificial Intelligence, Dongguk University, Seoul, South Korea（东国大学计算机科学与人工智能系）
