Benchmark RadarEgoMonth
EgoMonth benchmarks month-level egocentric video understanding with 300+ hours from 20 participants over 20-120 days, 1,443 QA pairs, and a 14-task framework across schema consolidation, episodic indexing, and cascading reasoning.
- Released
- 2026-08-13
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Existing long-video benchmarks lack inter-clip spatiotemporal continuity, so they cannot assess memory across days or weeks. EgoMonth provides a temporal-grounded evaluation for long-term memory in MLLMs, revealing that even top models remain far below human performance.
Motivation
Recent advances in Multimodal Large Language Models (MLLMs) have led to substantial progress in video understanding, accompanied by a growing number of long video benchmarks.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.