MM-Vet_FlashSloth-HD/metadata.json

{
  "dataset": "MM-Vet",
  "model_name": "FlashSloth",
  "model_links": [],
  "paper_title": "FlashSloth: Lightning Multimodal Large Language Models via Embedded Visual Compression",
  "paper_url": "https://arxiv.org/abs/2412.04317v1",
  "metrics": {
    "GPT-4 score": "41.9",
    "Params": "3.2B"
  },
  "table_metrics": {
    "GPT-4 score": "41.9",
    "Params": "3.2B"
  },
  "prompts": [
    "Given the following paper and codebase:\nPaper: FlashSloth: Lightning Multimodal Large Language Models via Embedded Visual Compression\nCodebase: Repository not available\n\nImprove the FlashSloth model on the MM-Vet dataset. The result should improve on the following metrics: {'GPT-4 score': '41.9', 'Params': '3.2B'}. You must use only the codebase provided."
  ]
}