AI Paper Atlas

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

Learning to Reason over Continuous Tokens with Reinforcement Learning

Yiran Zhao, Yuhui Xu, Doyen Sahoo, Caiming Xiong, Junnan Li

ICLR 2026 · Large language models › Reasoning and chain of thought · approach: Chain-of-thought training and analysis

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 learning to reason over continuous tokens in embedding space with reinforcement learning, aiming to keep the efficiency of latent reasoning while retaining clarity of discrete CoT.

Problem

Discrete chain-of-thought reasoning is computationally and memory expensive due to verbose intermediate steps, while existing latent reasoning in embedding space is efficient but often loses clarity and interpretability.

Approach

Performs reasoning over continuous token embeddings and trains the model with reinforcement learning so that latent reasoning steps are optimized directly for task success rather than imitating discrete CoT.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add GRPO reinforcement learning stage after cold start (ablation)HyRea (Qwen2.5-7B-Instruct) · MATH-500pass@1 accuracy: 71.8 -> 83.6Improved
add RL stage (ablation)HyRea (Qwen2.5-32B-Instruct) · MATH-500pass@1 accuracy: 79.8 -> 84.4Improved
use hybrid latent/explicit reasoning instead of SFT+RLQwen2.5-7B-Instruct · MATH-500pass@1 accuracy and tokens: 84.2 -> 83.6 accuracy; 698 -> 387 tokensMixed
use hybrid reasoning instead of pure latent reasoning (Coconut)Qwen2.5-7B-Instruct · MATH-500pass@1 accuracy: 70.4 -> 83.6Improved
select latent-replaced steps by low entropy instead of randomly (ablation)HyRea · MATH-500 tokensavg output tokens: 309 -> 287 tokens (accuracy 83.4 -> 83.6)Improved
select latent-replaced steps by low entropy instead of randomly (ablation)HyRea · AMC23pass@1 accuracy: 49.4 -> 48.2Hurt
increase number of latent tokens c from 1 to 8 (ablation)HyRea · math accuracyaccuracy: over 80% -> below 10%Hurt
apply HyRea-trained model to general tasks versus SFT+RLQwen2.5-7B-Instruct · MMLUaccuracy: 73.4 -> 68.6 (tokens 102 -> 53)Hurt
apply HyRea-trained model to GPQA versus SFT+RLQwen2.5-7B-Instruct · GPQAaccuracy: 29.8 -> 27.4 (tokens 1083 -> 685)Mixed

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.

Experimental setup

Builds on: chain-of-thought prompting

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