mmSSR Multi-modal SFT data selection method that first scales to million-level datapool, achieving 99.1% perf with 30% of LLaVA-OVSI. (in construction) Cream of the Crop: Harvesting Rich, Scalable and Transferable Multi-Modal Data for Instruction Fine-Tuning Paper • 2503.13383 • Published Mar 17 mengyaolyu/mmssr-7b-styler Updated Apr 3
Cream of the Crop: Harvesting Rich, Scalable and Transferable Multi-Modal Data for Instruction Fine-Tuning Paper • 2503.13383 • Published Mar 17
mmSSR Multi-modal SFT data selection method that first scales to million-level datapool, achieving 99.1% perf with 30% of LLaVA-OVSI. (in construction) Cream of the Crop: Harvesting Rich, Scalable and Transferable Multi-Modal Data for Instruction Fine-Tuning Paper • 2503.13383 • Published Mar 17 mengyaolyu/mmssr-7b-styler Updated Apr 3
Cream of the Crop: Harvesting Rich, Scalable and Transferable Multi-Modal Data for Instruction Fine-Tuning Paper • 2503.13383 • Published Mar 17