DynaTokens accepted at EMNLP: dynamic adaptation tokens for continual video-language learning
flosalim · x · 2026-09-22
The author announced that DynaTokens was accepted to the EMNLP 2026 main conference, tackling a pain point in continual learning for video-language models: adapting to new tasks without storing a separate set of adaptation tokens per task.
- Core idea: instead of storing task-specific prompts, the model learns to dynamically generate the right adaptation tokens when needed, controlling token dynamics.
- Implementation: frozen LLaMA-2-7B backbone + CLIP ViT-L/14, with LLaMA-Adapter for efficient video-language adaptation — no fine-tuning of the LLM backbone.
- Why it matters: avoids accumulating task-specific prompts over time, offering a new path for continual multi-task video-language adaptation.
Related event: DynaTokens Accepted to EMNLP 2026 for Continual Video-Language Adaptation(2 posts)→
More from Multimodal
- Dreamina 2.0 goes canvas-first: edit AI video locally instead of regenerating everything — eyishazyer · 2026-09-23
- Dreamina 2.0 bets on canvas-first editing as the new way to make AI video — eyishazyer · 2026-09-23
- fal's H3 Max generates 5 seconds of frontier-quality video in 3 seconds via full-stack optimization — gorkem · 2026-09-23
- China's AI micro-dramas grew 13x this year and are still hiring more people — Substantial-Fun9958 · 2026-09-23
- Orbis real-time video model hackathon draws 70+ builders and 20+ projects — _vztu · 2026-09-23
- Generating ads with Runway directly from ChatGPT is a neat workflow — tlakomy · 2026-09-23