MA-EgoQA: Question Answering over Egocentric Videos from Multiple Embodied Agents

Dimensions overall score 8.0

GitHub Code Pulse

No public code linked for this paper yet.

• The MA-EgoQA benchmark and code are publicly available for research use.technical 95%
• MA-EgoQA is the first benchmark designed to evaluate models on understanding multiple long-horizon egocentric videos from embodied agents.technical 95%
• MA-EgoQA contains 1.7k questions unique to multiple egocentric video streams across five reasoning categories.technical 90%
• Current approaches are unable to effectively handle multiple egocentric video streams.limitation 85%
• EgoMAS baseline model leverages shared memory and agent-wise dynamic retrieval for multi-agent egocentric video understanding.method 90%
• Effective compression and communication of high-volume individual sensory inputs is a key challenge in multi-agent egocentric video understanding.technical 80%
• Correct aggregation of multiple egocentric videos to construct system-level memory is a significant technical challenge.technical 80%
• Multi-agent embodied AI systems will require better communication between humans and AI agents in workplace or home settings.market 75%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card