Multimodal Weekly 73: Video-Language Models, Reasoning-Across-Time in Videos, Long-Horizon Multimodal Inference, and Scaling Vision Encoders

​​​​​In the 73rd session of Multimodal Weekly, we have four exciting presentations on video-language models, reasoning-across-time in videos, multimodality in long horizon causal inference, and scaling vision encoders for multimodal models.

​​​​✅ Peter Yu will present Espresso, a novel method that extracts and compresses spatial and temporal information in videos separately.

​​​​✅ Jr-Jen Chen will present ReXTime, a benchmark designed to rigorously test AI models' ability to perform temporal reasoning within video events.

​​​​✅ Zhuoyi Huang will present MARPLE, a benchmark for evaluating long-horizon inference capabilities using multi-modal evidence.

​​​​✅ Jieneng Chen will present the study on the analysis of redundancy concerning visual tokens and efficient training within large multimodal models.

​Join the Multimodal Minds community to connect with the speakers!

Multimodal Weekly is organized by Twelve Labs, a startup building multimodal foundation models for video understanding. Learn more about Twelve Labs here: https://twelvelabs.io/