VideoNexus framework released
The story so far
Hugging Face has released the VideoNexus framework, which separates video perception from evidence-driven planning reasoning. It uses perception sub-agents to return compact, temporally ordered evidence, achieving an accuracy of 88.6% on long video benchmarks, surpassing static ingestion and built-in agent processing, with API costs being only 44% of the latter.
AI-written from the reports · Updated 22 hours ago
Coverage timeline
Follow the coverage to see every side of the story.
- Hugging Face Daily PapersVideoNexus: Video Perception Beyond the Reasoning Context
VideoNexus separates detailed video perception from evidence-driven planning reasoning, using a perception sub-agent to return compact, temporally ordered evidence, achieving higher accuracy and lower cost in long video benchmarks, with an accuracy of 88.6% surpassing static ingestion and built-in agent processing, API cost being only 44% of it.
Heat over time
- Now
- 3
- Peak
- 419 Nov, 19:00
- 24h change
- –
Hover or tap the chart to see each hour; use the arrow keys on a keyboard.