Overview
Volcano employs a single LMM to generate initial responses, feedback, and revisions, as well as decisions to accept revisions. It follows a sequential procedure of an iterative critique-revision-decide loop.
Model details
Model type: Volcano is a multimodal self-feedback guided revision model that was trained using the vicuna model with visual instruction tuning data and multimodal feedback and revision data obtained through gpt-3.5-turbo, following the methodology of LLaVA.
Model date: Volcano-7b was trained in October 2023.
Paper or resources for more information:
Training dataset
- 274k Volcano-train data
- 558K filtered image-text pairs from LAION/CC/SBU, captioned by BLIP.
- 158K GPT-generated multimodal instruction-following data.
- 450K academic-task-oriented VQA data mixture.
- 40K ShareGPT data.