| Paper | Code | Accuracy | ModelName | ReleaseDate |
|---|---|---|---|---|
| Pushing the Frontier of Audiovisual Perception with Large-Scale Multimodal Correspondence Learning | ✓ Link | 78.9 | PE-AV Large (30 fps, zero-shot) (extra data) | 2025-12-22 |
| MoE-ViE: Mixture of Experts Vision Encoder for Efficient Image and Video Understanding | ✓ Link | 76.9 | MoE-ViE-H/14 | 2026-08-18 |
| Action100M: A Large-scale Video Action Dataset | ✓ Link | 64.8 | VL-JEPA Stage 3 ViT-L/256 (32 frames, zero-shot) (extra data) | 2026-01-15 |
| BT-Adapter: Video Conversation is Feasible Without Video Instruction Tuning | ✓ Link | 64.3 | BT-Adapter | 2023-09-27 |
| InternVideo: General Video Foundation Models via Generative and Discriminative Learning | ✓ Link | 64.25 | InternVideo (zero-shot) | 2022-12-06 |