Rethinking Visual Intelligence: Insights from Video Pretraining
This work addresses the challenge of building more capable visual foundation models for AI systems, though it appears incremental as it builds on existing pretraining paradigms.
The paper tackled the problem of limited adaptability and sample efficiency in visual models compared to language models by investigating Video Diffusion Models (VDMs) pretrained on spatiotemporal data, finding that VDMs demonstrated higher data efficiency across multiple benchmarks such as ARC-AGI and visual games.
Large language models (LLMs) have demonstrated that large-scale pretraining enables systems to adapt rapidly to new problems with little supervision in the language domain. This success, however, has not translated as effectively to the visual domain, where models, including LLMs, continue to struggle with compositional understanding, sample efficiency, and general-purpose problem-solving. We investigate Video Diffusion Models (VDMs) as a promising direction for bridging this gap. Pretraining on spatiotemporal data endows these models with strong inductive biases for structure and dynamics, which we hypothesize can support broad task adaptability. To test this, we design a controlled evaluation in which both a pretrained LLM and a pretrained VDM are equipped with lightweight adapters and presented with tasks in their natural modalities. Across benchmarks including ARC-AGI, ConceptARC, visual games, route planning, and cellular automata, VDMs demonstrate higher data efficiency than their language counterparts. Taken together, our results indicate that video pretraining offers inductive biases that support progress toward visual foundation models.