SAGE: Training Smart Any-Horizon Agents for Long Video Reasoning with Reinforcement Learning
Open paper
Benchmark
Model
Rank
Results
video-question-answering-on-videommmu
SAGE-Flash (SAGE-MM: Qwen3-VL-8B-Instruct [+SFT] [+RL])
#3
Accuracy: 68.1
video-question-answering-on-video-mmmu
SAGE-Flash
#17
Accuracy: 68.1
Rank counts only results with a code link.