CLEVRER: CoLlision Events for Video REpresentation and Reasoning

TL;DR

CLEVRER introduces a diagnostic video dataset designed to evaluate models on causal reasoning tasks involving object collision events, highlighting the gap between perception and causal understanding in current models.

Contribution

The paper presents CLEVRER, a new dataset for systematic evaluation of causal reasoning in videos, and analyzes the performance gap of state-of-the-art models on causal tasks.

Findings

Models excel at perception-based tasks but struggle with causal reasoning.

Explicit symbolic models improve causal reasoning performance.

Current models lack integrated perception and causal understanding capabilities.

Abstract

The ability to reason about temporal and causal events from videos lies at the core of human intelligence. Most video reasoning benchmarks, however, focus on pattern recognition from complex visual and language input, instead of on causal structure. We study the complementary problem, exploring the temporal and causal structures behind videos of objects with simple visual appearance. To this end, we introduce the CoLlision Events for Video REpresentation and Reasoning (CLEVRER), a diagnostic video dataset for systematic evaluation of computational models on a wide range of reasoning tasks. Motivated by the theory of human casual judgment, CLEVRER includes four types of questions: descriptive (e.g., "what color"), explanatory ("what is responsible for"), predictive ("what will happen next"), and counterfactual ("what if"). We evaluate various state-of-the-art models for visual reasoning on our benchmark. While these models thrive on the perception-based task (descriptive), they perform poorly on the causal tasks (explanatory, predictive and counterfactual), suggesting that a principled approach for causal reasoning should incorporate the capability of both perceiving complex visual and language inputs, and understanding the underlying dynamics and causal relations. We also study an oracle model that explicitly combines these components via symbolic representations.