$C^3$: Compositional Counterfactual Contrastive Learning for Video-grounded Dialogues

TL;DR

This paper introduces $C^3$, a novel contrastive learning method for video-grounded dialogues that enhances multimodal reasoning and generalization by leveraging counterfactual samples and object/action-level variance.

Contribution

It proposes a new compositional counterfactual contrastive learning framework that improves video-grounded dialogue systems by focusing on hidden state representations and counterfactual sampling.

Findings

Achieved performance gains on AVSD benchmark

Improved grounding of video and dialogue context

Enhanced multimodal reasoning capabilities

Abstract

Video-grounded dialogue systems aim to integrate video understanding and dialogue understanding to generate responses that are relevant to both the dialogue and video context. Most existing approaches employ deep learning models and have achieved remarkable performance, given the relatively small datasets available. However, the results are partly accomplished by exploiting biases in the datasets rather than developing multimodal reasoning, resulting in limited generalization. In this paper, we propose a novel approach of Compositional Counterfactual Contrastive Learning ($C^3$) to develop contrastive training between factual and counterfactual samples in video-grounded dialogues. Specifically, we design factual/counterfactual sampling based on the temporal steps in videos and tokens in dialogues and propose contrastive loss functions that exploit object-level or action-level variance. Different from prior approaches, we focus on contrastive hidden state representations among compositional output tokens to optimize the representation space in a generation setting. We achieved promising performance gains on the Audio-Visual Scene-Aware Dialogues (AVSD) benchmark and showed the benefits of our approach in grounding video and dialogue context.