FALCONEye: Finding Answers and Localizing Content in ONE-hour-long videos with multi-modal LLMs

Carlos Plou, Cesar Borja, Ruben Martinez-Cantin, Ana C. Murillo,

🌍 Homepage | 🤗 Benchmark | 📝 ArXiv

🔔 News:

• 🔜 Code will be released upon acceptance.
• 🥳 3/26/2025: We have released the FALCON-Bench and Paper! 🔥

Description

This repo contains the code presented in the paper "FALCONEye: Finding Answers and Localizing Content in ONE-hour-long videos with multi-modal LLMs". FALCONEye code was built under the lmms-eval framework.

📝 Citation

@article{plou2025falconeyefindinganswerslocalizing,
      title={FALCONEye: Finding Answers and Localizing Content in ONE-hour-long videos with multi-modal LLMs}, 
      author={Carlos Plou and Cesar Borja and Ruben Martinez-Cantin and Ana C. Murillo},
      year={2025},
      eprint={2503.19850},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2503.19850},
}