AI Paper Atlas

Research map / Large language models / Reasoning and chain of thought

Beyond Logits: Metastable Latent Dynamics for Sample-Efficient Best-of-N Selection in LLMs

Xinrong Li, Zidong Zhou, Keyu Shen, Wenhao Zhou, Shangqi Guo

ICML 2026 · Large language models › Reasoning and chain of thought · approach: Test-time compute scaling

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes a best-of-N selection method for LLM reasoning based on metastable latent dynamics rather than output logits or sample agreement, aiming for sample-efficient verification.

Problem

Best-of-N selection for LLM reasoning needs reliable verification; logit-confidence proxies suffer calibration collapse and sample-agreement proxies are costly or brittle.

Approach

Analyzes latent (hidden-state) dynamics of candidate trajectories and uses metastability as a verification signal for selecting among N samples, instead of relying on output logit confidence or self-consistency.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
select by LVE instead of Self-Consistency at N=3LVE · MATH500 (Qwen, N=3)Pass@1 accuracy (%): 84.2 -> 87.4Improved
select by LogNorm-LVE instead of Self-Consistency at N=10LogNorm-LVE · MATH500 (Qwen, N=10)Pass@1 accuracy (%): 87.6 -> 92.2Improved
select by LogNorm-LVE instead of Self-Consistency at N=5LogNorm-LVE · MATH500 (Qwen, N=5)Pass@1 accuracy (%): 87.0 -> 89.0Improved
select by LogNorm-LVE instead of Self-Consistency at N=15LogNorm-LVE · MATH500 (Qwen, N=15)Pass@1 accuracy (%): 88.0 -> 91.4Improved
select by Self-Certainty instead of random choiceSelf-Certainty · MATH500 (Qwen, N=10)Pass@1 accuracy (%): 81.4 -> 70.2Hurt
add length normalization (LogNorm-LVE vs raw LVE) (ablation)LVE · MATH500 (Qwen, N=10)Pass@1 accuracy (%): 90.4 -> 92.2Improved
add length normalization (LogNorm-LVE vs raw LVE) (ablation)LVE · MATH500 (Llama, N=10)Pass@1 accuracy (%): 85.4 -> 85.0No clear effect
add length normalization (LogNorm-LVE vs raw LVE) (ablation)LVE · GPQA (Qwen, N=10)Pass@1 accuracy (%): 55.6 -> 53.5Hurt
add raw LVE to a length-only feature for ranking candidates (ablation)length-only classifier · MATH500 (Qwen)AUROC: 0.517 -> 0.626Improved
score with LVE hidden-state replay instead of logits-based scoringLVE · MATH500 (Llama, N=10)scoring latency (ms): 1257 -> 870Improved
select by LVE instead of random choiceLVE · Average of four benchmarks (Qwen, N=10)Pass@1 accuracy (%): 44.4 -> 56.5Improved
select by LVE instead of Self-Consistency on GPQALVE · GPQA (Qwen, N=10)Pass@1 accuracy (%): 52.5 -> 55.6No clear effect

Surprising findings

Experimental setup

Baselines: logit confidence sample agreement

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy