一种用于对抗具身推理中时空幻觉的渐进训练策略
A Progressive Training Strategy for Embodied Vision-Language Models to Mitigate Spatio-Temporal Hallucinations
AI总结：
本文提出一种渐进训练框架，通过构建CoT数据集和弱标签数据提升视觉语言模型的时空推理能力，有效缩小了正反向性能差距。
摘要：
视觉语言模型（VLMs）在静态图像理解上取得了显著进展，但在时空推理方面仍面临关键挑战。主要瓶颈是'多图像推理幻觉'，即正向与反向时间查询之间性能大幅下降，表明依赖于表面捷径而非真实因果理解。为此，我们首先开发了一个新的链式思维（CoT）数据集，将复杂推理分解为详细的时空步骤和确定性判断。在此基础上，我们提出了一种渐进训练框架：首先在CoT数据集上进行监督预训练以建立逻辑结构，随后通过可扩展的弱标签数据进行微调以实现更广泛的一般化。实验表明，该方法不仅提高了基础模型的准确性，还将正反向性能差距从超过70%降至仅6.53%。这证实了该方法能够发展出真实的动态推理能力，并减少当前VLMs固有的时间偏差。
英文摘要：
Vision-Language Models (VLMs) have made significant strides in static image understanding but continue to face critical hurdles in spatiotemporal reasoning. A major bottleneck is "multi-image reasoning hallucination", where a large performance drop between forward and reverse temporal queries reveals a dependence on superficial shortcuts instead of state-based understanding. To mitigate this, we first develop a new Chain-of-Thought (CoT) dataset that decomposes intricate reasoning into detailed spatiotemporal steps and definitive judgments. Building on this, we present a progressive training framework: it initiates with supervised pre-training on our CoT dataset to instill logical structures, followed by fine-tuning with scalable weakly-labeled data for broader generalization. Our experiments demonstrate that this approach not only improves backbone accuracy but also reduces the forward-backward performance gap from over 70% to only 6.53%. This shows that the method strengthens dynamic reasoning and reduces the inherent temporal biases of current VLMs.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
