Research Papers

• QMAVIS: Long Video-Audio Understanding using Fusion of Large Multimodal Models FUSION 2025

  Co-authored research paper on long video-audio understanding via late-fusion of LMMs/LLMs + ASR; reported +38.75% improvement on VideoMME (w/ subtitles) vs baselines like VideoLLaMA2 and InternVL2. Accepted at the 28th International Conference on Information Fusion (FUSION 2025), Rio de Janeiro, Brazil.

  View on arXiv (arXiv:2601.06573)