arXiv cs.CL· Wei Xia, Yehui Tang, Zhi-Hong Deng·· 2 天前AI 评分29
EMO：交替上下文优化与 logits 引导的闭环测试时计算框架
EMO: A Closed-Loop Test-Time Computation Framework via Alternating Context Optimization and Logits Steering
AI 导读
EMO 是一个无需训练的测试时计算框架，受 EM 算法启发，交替进行问题特定的上下文估计（E 步）与分布精炼（M 步），通过外部评判模型的文本梯度修正系统上下文，并利用上下文引起的 logit 位移构建置信度约束的 logit 修正。
正文 · 原文
Abstract:Test-time context optimization adapts frozen language models through textual feedback, but ordinary conditional generation provides no separate control over how strongly to reinforce the distributional changes induced by a context update. We introduce EMO, a training-free framework inspired by expectation--maximization that alternates question-specific context estimation (E-step) and distribution refinement (M-step). The E-step revises a system context using textual gradients from an external judge without access to gold answers. The M-step contrasts next-token distributions under the revised and preceding contexts at a shared prefix to construct a confidence-constrained logit correction. Feedback on the steered response drives the next context update, coupling adaptation across rounds with intervention within each generation. Theoretically, we connect parameter and context optimization through a shared target-KL objective and additive logit representations, and derive the steering vector from context-induced logit displacements. Under context-improvement and faithful-steering assumptions, we prove a one-round decrease in target KL divergence for sufficiently small steering strength. Under greedy decoding across eight models and six benchmarks, EMO improves the zero-shot baseline in 39 of 40 aggregated settings, averaging $+9.10$ percentage points. With the same judge, it adds $+1.53$ points on average over context-only optimization (TextGrad), with consistent MMLU-Pro gains across models. At matched target-model generation budgets, accuracy advantages over best-of-$N$ are strongest on Llama, while oracle coverage is higher in most comparisons despite fewer candidates. The coverage--selection gap motivates allocating test-time computation jointly to candidate generation and verification.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.22160 [cs.CL] |
| (or arXiv:2610.22160v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.22160 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Wei Xia [view email]
[v1]
Thu, 15 Oct 2026 16:23:55 UTC (437 KB)
来源：arXiv cs.CL · arxiv.org