Parallel Autoregressive Decoding Framework Boosts Dense Video Captioning

A forthcoming ECCV2026 paper introduces a parallel autoregressive decoding framework for omni-modal dense video captioning. Developed by NUS, the method significantly improves generation efficiency by leveraging the independence of temporally distant video events.

2026-07-08 ~ 2026-07-09 · 2 related posts

• NUS Proposes Parallel Autoregressive Decoding for Efficient Dense Video Captioning — NationalUniversityofSingapore · 2026-07-08
• ECCV2026: Parallel Autoregressive Decoding for Omni-modal Video Captioning — MikeShou1 · 2026-07-09