跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2611.24032 · v1

Test-Time Symmetrization: Scaling Laws for Model-Agnostic Learning under Structure

首次提交 2026-11-18 16:20
版本更新 2026-11-18 16:20
雷达观测 2026-11-19 19:07
arXiv ↗PDF ↗

摘要

Test-time compute has emerged as a powerful way to improve the accuracy of language models at inference time, with scaling laws quantifying how accuracy improves with additional forward passes. Existing scaling laws, however, focus on repeated sampling from a fixed prompt, even though many tasks admit multiple equivalent representations of the same underlying problem. A graph does not depend on how its nodes are numbered, and a program does not depend on the names of its variables, yet language models can change their answers under such rewritings. In this paper, we study test-time symmetrization, which uses forward passes on rewritings of the input and aggregates the resulting answers over the group GG of rewritings. We prove that no amount of sampling and repeated decoding from a single representation can replace rewriting, while only O(log⁡∣G∣)O(\log |G|) random rewritings suffice to recover the information provided by all ∣G∣|G| rewritings. The same set of O(log⁡∣G∣)O(\log |G|) rewritings provides simultaneous approximation across models, decoding rules, and sampled reasoning traces, and the logarithmic scaling is optimal in general. We validate the theory with experiments on Qwen2.5-7B, Phi-4-mini, Llama-3.1-8B, and Qwen2.5-14B, with chain-of-thought reasoning, on graph reasoning, logical deduction, and multiple-choice question answering. At the same number of forward passes, spending them on rewritings of the input is substantially more accurate than spending them on repeated sampling from the original prompt whenever the model's answer depends on how the input is written, and the number of rewritings needed grows only logarithmically with the number of possible rewritings, as the theory predicts. Our results provide a principled way to enforce symmetries at inference time when they cannot be built into the model, which we call model-agnostic symmetrization.

作者备注:38 pages, 13 figures

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

目前只收录了一个版本。收录下一版后,可以对照标题、作者、分类和摘要。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v1 · 2026-11-18 16:20 · 最新观测版本

Test-Time Symmetrization: Scaling Laws for Model-Agnostic Learning under Structure

Test-time compute has emerged as a powerful way to improve the accuracy of language models at inference time, with scaling laws quantifying how accuracy improves with additional forward passes. Existing scaling laws, however, focus on repeated sampling from a fixed prompt, even though many tasks admit multiple equivalent representations of the same underlying problem. A graph does not depend on how its nodes are numbered, and a program does not depend on the names of its variables, yet language models can change their answers under such rewritings. In this paper, we study test-time symmetrization, which uses forward passes on rewritings of the input and aggregates the resulting answers over the group GG of rewritings. We prove that no amount of sampling and repeated decoding from a single representation can replace rewriting, while only O(log⁡∣G∣)O(\log |G|) random rewritings suffice to recover the information provided by all ∣G∣|G| rewritings. The same set of O(log⁡∣G∣)O(\log |G|) rewritings provides simultaneous approximation across models, decoding rules, and sampled reasoning traces, and the logarithmic scaling is optimal in general. We validate the theory with experiments on Qwen2.5-7B, Phi-4-mini, Llama-3.1-8B, and Qwen2.5-14B, with chain-of-thought reasoning, on graph reasoning, logical deduction, and multiple-choice question answering. At the same number of forward passes, spending them on rewritings of the input is substantially more accurate than spending them on repeated sampling from the original prompt whenever the model's answer depends on how the input is written, and the number of rewritings needed grows only logarithmically with the number of possible rewritings, as the theory predicts. Our results provide a principled way to enforce symmetries at inference time when they cannot be built into the model, which we call model-agnostic symmetrization.

查看此版本解读 →查看此版本原文 ↗