---
title: "VBVR-Pro：面向原生视觉推理的可扩展可验证测试套件"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2608.26105v1"
permalink: "https://feed.astralor.com/items/itm_615e5519323f532c262da8b4"
published: "2026-08-26T17:59:51.000Z"
score: 78
channel: "ai"
tags: ["VBVR-Pro","原生视觉推理","可验证奖励","强化学习","视觉生成"]
---

# VBVR-Pro：面向原生视觉推理的可扩展可验证测试套件

## AI 点评

> 这篇论文把“视觉生成即推理”从口号变成了可实验的体系：既有任务扩张，又有可靠奖励，还做了模态消融。对做视觉强化学习、多模态评测和生成式推理的研究者尤其值得细读，其发布的全套资源可能成为新的社区基座。

## 速览

VBVR-Pro 是面向原生视觉推理的闭环测试套件，将视觉生成本身作为推理媒介。它用 300 个程序化任务实现可扩展训练，模型在七个外部基准上表现强劲；提供基于确定性规则的可验证奖励评分器，克服 VLM-as-judge 的通病，并支撑大规模多任务强化学习；还支持对 30 余种图像、视频与交错生成器的受控模态对比，发现视频生成最擅长时空状态追踪，交错生成更省算力。数据、模型、评分器与代码全部开源。

## 关键点

- 300 个程序化生成任务构成可扩展训练空间，模型可迁移至 RISE-Video、MME-CoF-Pro、BabyVision 等 7 个外部基准。
- 提出基于确定性规则的可验证奖励评分器，优于流行的 VLM-as-judge 模式，并可作为大规模多任务强化学习的稳定奖励信号。
- 对 30 余种图像、视频和交错生成器进行受控模态研究，发现视频生成在持续时空状态追踪上最强，交错生成则更高效。
- 全部数据、模型、评分器和代码已开源，便于复现与后续扩展。

## 全文

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.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_615e5519323f532c262da8b4