Boosting Omni-Modal Language Models: Staged Post-Training with Visually Debiased Evaluation
Open paper
Benchmark
Model
Rank
Results
video-question-answering-on-videoholmes
Qwen2.5-Omni-3B (Stage 2: Mixed-Modality RLVR)
–
Accuracy: 47.07
Rank counts only results with a code link.