CoBa: Cost-Effective Test-Time Scaling via Compute-Balanced Routing
2026-08-07 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors address the problem of using limited computing power optimally when a system tries to solve problems by generating answers and checking them. They present CoBa, a method that first generates a small number of candidate answers, quickly verifies them, and then applies more thorough checks only to uncertain or promising candidates. Their approach achieves accuracy comparable to more expensive methods but uses much less computing. This shows that intelligently deciding where to spend computing resources during problem-solving can improve efficiency without losing accuracy.
test-time scalingcompute allocationchain of thoughtverificationparameter-weighted tokensmacro accuracybootstrap testrouting policymajority voting
Authors
Yan Zhou, Yue Ouyang, Kaiyang Zheng, Suncheng Xiang
Abstract
Test-time scaling is often implemented by spending more compute along one axis: sampling more solutions, extending a chain of thought, or applying a stronger evaluator. Under a fixed inference budget, these choices compete. This paper formulates test-time reasoning as a compute-allocation problem in which a system must decide whether the next unit of compute should be spent on generation, verification, or stopping. We introduce CoBa, a compute-balanced routing policy that first obtains a small set of candidates, applies cheap verification broadly, and routes uncertain or high-value candidates to stronger verification. On 3,129 example-generator evaluations spanning MATH-500, AIME 2024/2025, AMC 2023, and procedural symbolic reasoning, CoBa-Routed-Strong reaches 85.13% macro accuracy, statistically matching a self-evaluation weighted-voting proxy at 85.20% while using 49.1% fewer parameter-weighted tokens. It also matches best-of-16 majority voting within 0.01 macro-accuracy points while using 58.9% fewer parameter-weighted tokens; paired tests retain a small best-of-16 edge at substantially higher cost. Paired bootstrap tests show significant gains over single-sample decoding, while the remaining gap to the pool oracle exposes headroom for sharper routing. For local reasoning systems, test-time scaling becomes a question of where the next computation is most valuable.