AI Paper Atlas

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

Once-More: Continuous Self-Correction for Large Language Models via Perplexity-Guided Intervention

Jiaxun Gao, Him Wai Ng, Z. Jane Wang

ICLR 2026 · Large language models › Reasoning and chain of thought · approach: Routing, pruning and distillation for reasoning

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 Once-More, a method for continuous self-correction of LLMs during long generation, guided by perplexity-based intervention at inference time.

Problem

Compounding errors during long LLM text generation, where early mistakes propagate into drift, faulty reasoning, or repetition, and self-correction behaviour is unpredictable at inference time.

Approach

Uses perplexity as a guide to intervene during generation so the model continuously self-corrects rather than relying on unpredictable emergent self-correction from scale; details beyond the abstract are not available.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
apply Once-More continuous self-correctionQwen3-14B · AIME24accuracy: 26.7 -> 36.7Improved
apply Once-MoreQwen3-4B · LiveBench (reasoning)accuracy: 20.3 -> 33.0Improved
apply Once-MoreQwen3-4B · GPQA Diamondaccuracy: CRITIC 48.9 vs Once-More 47.5Mixed
apply Self-RefineQwen3-8B · AIME25accuracy: 16.7 -> 15.5Hurt
add verifier feedback on top of logit redistribution (ablation)Once-More Qwen3-14B · GPQA Diamondaccuracy: 48.4 -> 55.6Improved
add logit redistribution on top of verifier feedback (ablation)Once-More Qwen3-14B · AIME24accuracy: 33.3 -> 36.7Improved
increase unit length from 1 to 128 sentences (ablation)Once-More Qwen3-14B · AIME24accuracy: 36.6 -> 26.7Hurt
vary unit length among 1, 2, 4 sentences (ablation)Once-More Qwen3-14B · AIME24accuracy: 36.6 / 34.3 / 35.5No clear effect
increase suppression strength alpha from 0.1 to 1 (ablation)Once-More Qwen3-14B · AIME24accuracy: 30.0 -> 36.6Improved
use position-only suppression (tau=0.01) instead of distance decay tau=1 (ablation)Once-More Qwen3-14B · AIME24accuracy: 36.6 -> 23.3Hurt
use stronger 14B verifier with 4B producer instead of 4B verifier (ablation)Once-More Qwen3-4B · AIME24accuracy: 16.7 -> 33.3Improved
apply Once-More vs token usage of Self-RefineQwen3-14B · AIMEtoken usage: 17-21% fewer tokensImproved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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