# ChronoVision：通过潜在状态重建实现时序推理

- 来源：HuggingFace Daily Papers（社区热门论文）
- 发布时间：2026-08-06 08:00
- AIHOT 分数：55
- AIHOT 链接：https://aihot.virxact.com/items/cmsicrk3r19c9ronkej7t65qg
- 原文链接：https://arxiv.org/abs/2608.05631

## AI 摘要

ChronoVision 提出多模态框架，通过监督微调中的 Reconstructive Visual Head 预测最终变换状态的潜在表征，并借助 ROI Attention Locating 模块聚焦关键视觉证据，后训练阶段采用带隐式过程对齐的强化学习。在 Vbvr-VQA 上取得 74.8% 域内和 71.6% 域外准确率，在 IntPhys2 上达到 55.0%。

## 正文

Multimodal large language models excel at passive perception but struggle with complex visual cognitive tasks requiring multi-step temporal reasoning. This degradation largely stems from the inherent ambiguity of language-based reasoning, which often fails to accurately articulate continuous visual transformations. To address this, we propose ChronoVision, a multimodal framework designed to align visual logic with latent imagery. During supervised fine-tuning, a Reconstructive Visual Head predicts the latent representation of the final transformed state, while an ROI Attention Locating module focuses the model on key visual evidence via semantic span queries. In post-training, we apply reinforcement learning with an implicit process grounding mechanism, guided by a composite reward function that evaluates outcome correctness, latent process alignment, and unsupervised visual focus. Furthermore, we introduce Vbvr-VQA, a novel dataset that evaluates temporal tracking by reformulating video reasoning into a strict image-ordering task. Experiments demonstrate that ChronoVision achieves state-of-the-art performance on Vbvr-VQA with 74.8% in-domain and 71.6% out-of-domain accuracy, alongside a strong 55.0% accuracy on IntPhys2, a highly challenging cross-domain benchmark.