VBVR-Pro: A Scalable and Verifiable Suite for Native Visual Reasoning

Hugging Face Daily Papers Papers

Summary

VBVR-Pro introduces a closed-loop testbed for scalable and verifiable native visual reasoning through generation, featuring task scaling, verifiable rewards, and mechanism studies across diverse visual substrates.

Native visual reasoning treats visual generation as the medium of reasoning itself: visual states (i.e. images and videos) are not merely inputs to be understood or outputs to be rendered, but first-class substrates for problem solving beyond language. Yet progress remains bottlenecked by the lack of scalable training tasks, reliable feedback, and controlled comparisons across generative substrates. In this work, we introduce VBVR-Pro, a closed-loop testbed that makes native visual reasoning through generation trainable, verifiable, optimizable, and experimentally controllable. 1) Task scaling. VBVR-Pro turns visual reasoning into a controlled task space of 300 procedurally generated tasks. Models trained on VBVR-Pro show strong transfer beyond the proposed suite across seven external visual reasoning benchmarks such as RISE-Video, MME-CoF-Pro, and BabyVision. 2) Verifiable rewards. VBVR-Pro provides verifiable reward scorers for task-grounded evaluation. Through a systematic study of leading MLLMs as judges, we identify recurring failure modes of the prevalent VLM-as-a-judge paradigm. In contrast, the proposed scorers are grounded in deterministic, task-specific rules, achieve fine-grained alignment with human judgments. Importantly, they serve as reliable reward signals for large-scale multi-task reinforcement learning and demonstrate stronger post-RL performance across visual reasoning tasks. 3) Mechanism study. VBVR-Pro enables controlled modality studies across more than 30 image, video, and interleaved generators. Our analysis shows that video generation remains strongest for tasks requiring persistent spatiotemporal state tracking, while interleaved generation provides a compute-efficient alternative. Critically, ablations and probing suggest the presence of vision-native trajectories that are crucial to visual reasoning. We release all data, models, scorers, and code.
Original Article
View Cached Full Text

Cached at: 08/27/26, 03:18 AM

Paper page - VBVR-Pro: A Scalable and Verifiable Suite for Native Visual Reasoning

Source: https://huggingface.co/papers/2608.26105 Published on Aug 26

#3 Paper of the day Authors:

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

Abstract

VBBR-Pro introduces a closed-loop testbed that enables scalable, verifiable, and controllable native visual reasoning through generation across diverse visual substrates.

Native visual reasoningtreatsvisual generationas the medium of reasoning itself: visual states (i.e. images and videos) are not merely inputs to be understood or outputs to be rendered, but first-class substrates for problem solving beyond language. Yet progress remains bottlenecked by the lack of scalable training tasks, reliable feedback, and controlled comparisons across generative substrates. In this work, we introduce VBVR-Pro, a closed-loop testbed that makesnative visual reasoningthrough generation trainable, verifiable, optimizable, and experimentally controllable. 1) Task scaling. VBVR-Pro turns visual reasoning into a controlled task space of 300 procedurally generated tasks. Models trained on VBVR-Pro show strong transfer beyond the proposed suite across seven external visual reasoning benchmarks such as RISE-Video, MME-CoF-Pro, and BabyVision. 2) Verifiable rewards. VBVR-Pro providesverifiable reward scorersfor task-grounded evaluation. Through a systematic study of leadingMLLMsas judges, we identify recurring failure modes of the prevalentVLM-as-a-judgeparadigm. In contrast, the proposed scorers are grounded in deterministic, task-specific rules, achieve fine-grained alignment with human judgments. Importantly, they serve as reliable reward signals for large-scalemulti-task reinforcement learningand demonstrate stronger post-RL performance across visual reasoning tasks. 3) Mechanism study. VBVR-Pro enables controlled modality studies across more than 30 image, video, and interleaved generators. Our analysis shows that video generation remains strongest for tasks requiring persistent spatiotemporal state tracking, whileinterleaved generationprovides a compute-efficient alternative. Critically, ablations and probing suggest the presence ofvision-native trajectoriesthat are crucial to visual reasoning. We release all data, models, scorers, and code.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2608\.26105

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/2608.26105 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2608.26105 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2608.26105 in a Space README.md to link it from this page.

Collections including this paper1

Similar Articles

A Very Big Video Reasoning Suite

Papers with Code Trending

This paper introduces the Very Big Video Reasoning (VBVR) dataset and benchmark, a large-scale resource with over one million video clips across 200 reasoning tasks, enabling systematic study of spatiotemporal reasoning and showing early signs of emergent generalization.

UniVR: Thinking in Visual Space for Unified Visual Reasoning

Hugging Face Daily Papers

UniVR introduces VR-GRPO, a reinforcement learning paradigm for unified visual reasoning, learning complex reasoning and physical dynamics from pure visual demonstrations, achieving up to 25% improvement on the VR-X benchmark.

Unlocking Complex Visual Generation via Closed-Loop Verified Reasoning

Hugging Face Daily Papers

Introduces CLVR (Closed-Loop Visual Reasoning), a framework that reformulates text-to-image generation from a single-step process into a closed-loop, multi-step visual reasoning approach using a VLM controller and diffusion models, achieving improved performance on compositional prompts.