VeriEvol: Scaling Multimodal Mathematical Reasoning via Verifiable Evol-Instruct

Hugging Face Daily Papers Papers
multimodal mathematical-reasoning reinforcement-learning verifiable-evol-instruct visual-reasoning data-scaling hypothesis-testing

Summary

VeriEvol is a novel framework for scaling reinforcement learning in visual mathematical reasoning by ensuring reliable reward labels through a two-axis approach separating prompt difficulty from answer reliability, using evolutionary operators and hypothesis-testing verification. It achieves significant accuracy gains on a five-benchmark visual-math suite.

Scaling reinforcement learning for visual mathematical reasoning requires more than generating harder questions: as data volume grows, the reward labels themselves must remain reliable. Yet existing data pipelines scale supervision while trusting the labeller, and policy-side methods assume the underlying answers are already correct. We instead treat scaling as a verifiable data-construction problem and decouple two axes before any policy update: prompt difficulty, expanded by route-specific evolution operators, and answer reliability, enforced by offline hypothesis-test falsification. We instantiate this as VeriEvol, an iterative framework with two extensible components: a type-aware evolution module that rewrites low-difficulty image-question seeds into harder, image-grounded prompts; and HTV-Agent, a verifier that accepts an answer only after multi-source counter-evidence has failed to refute it. The resulting verified data scales in volume, extends by adding evolution routes or verifier channels, and plugs directly into existing GRPO-style RL recipes. On a five-benchmark visual-math suite, scaling evolved SFT data from 10K to 250K samples raises the mean accuracy from 35.42 to 54.73; then, with backbone, SFT initialization, and GRPO recipe held fixed, VeriEvol adds a cumulative +3.88 over an un-evolved RL baseline, of which +1.82 comes from evolved prompts and +2.06 from the HTV-Agent verifier. We release the prompts, data, models, code, and the full verifier trace of every sample, so that downstream work can scale and audit the pipeline rather than only inspect its outputs.
Original Article
View Cached Full Text

Cached at: 06/25/26, 05:17 AM

Paper page - VeriEvol: Scaling Multimodal Mathematical Reasoning via Verifiable Evol-Instruct

Source: https://huggingface.co/papers/2606.23543

Abstract

A novel framework called VeriEvol is introduced that addresses the challenge of scaling reinforcement learning for visual mathematical reasoning by ensuring reliable reward labels through a two-axis approach that separates prompt difficulty from answer reliability, utilizing evolutionary operators and hypothesis testing verification to improve model performance and transparency.

Scalingreinforcement learningforvisual mathematical reasoningrequires more than generating harder questions: as data volume grows, thereward labelsthemselves must remain reliable. Yet existing data pipelines scale supervision while trusting the labeller, and policy-side methods assume the underlying answers are already correct. We instead treat scaling as averifiable data-constructionproblem and decouple two axes before any policy update:prompt difficulty, expanded by route-specificevolution operators, andanswer reliability, enforced by offlinehypothesis-test falsification. We instantiate this as VeriEvol, aniterative frameworkwith two extensible components: atype-aware evolution modulethat rewrites low-difficulty image-question seeds into harder, image-grounded prompts; andHTV-Agent, averifierthat accepts an answer only after multi-source counter-evidence has failed to refute it. The resulting verified data scales in volume, extends by adding evolution routes orverifierchannels, and plugs directly into existingGRPO-style RLrecipes. On a five-benchmark visual-math suite, scaling evolvedSFT datafrom 10K to 250K samples raises the mean accuracy from 35.42 to 54.73; then, with backbone, SFT initialization, and GRPO recipe held fixed, VeriEvol adds a cumulative +3.88 over an un-evolved RL baseline, of which +1.82 comes from evolved prompts and +2.06 from theHTV-Agentverifier. We release the prompts, data, models, code, and the fullverifiertrace of every sample, so that downstream work can scale and audit the pipeline rather than only inspect its outputs.

View arXiv pageView PDFProject pageGitHub2Add to collection

Get this paper in your agent:

hf papers read 2606\.23543

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2606.23543 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2606.23543 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2606.23543 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

REVES: REvision and VErification--Augmented Training for Test-Time Scaling

Hugging Face Daily Papers

Proposes REVES, a two-stage iterative framework that alternates between data augmentation and policy optimization to improve LLM reasoning by leveraging intermediate correction steps, achieving superior performance on coding benchmarks and constraint satisfaction problems.

OmniThoughtVis: A Scalable Distillation Pipeline for Deployable Multimodal Reasoning Models

arXiv cs.CL

This paper introduces OmniThoughtVis, a scalable pipeline for distilling multimodal reasoning capabilities from large teacher models to smaller, deployment-oriented MLLMs. The method uses curated chain-of-thought data to significantly improve reasoning performance on benchmarks like MathVerse and MMMU-Pro for models ranging from 2B to 8B parameters.