Qwen3-4B MedMCQA 10K checkpoints: SFT, DFT, ASFT, and future S³FT variants.
Seongryong Jung
SeongryongJung
·
AI & ML interests
Post-training, Knowledge Distillation, Self-Evolving AI
Recent Activity
updated a dataset 4 days ago
SeongryongJung/powerplant-shortqa-rows-4001-5000 published a dataset 4 days ago
SeongryongJung/powerplant-shortqa-rows-4001-5000 updated a collection 5 days ago
S³FT Medical Adaptation: Qwen3-4B on MedMCQAOrganizations
models 25
SeongryongJung/llama2-7b-medmcqa-30k-s3ft-base-rkl-beta0.3-lora
Updated • 19
SeongryongJung/llama2-7b-medmcqa-30k-s3ft-adaptive-hardgrow-base-rkl-beta0.3-lora
Updated • 23
SeongryongJung/llama2-7b-medmcqa-10k-s3ft-adaptive-hardgrow-base-rkl-beta0.3-lora
Updated • 23
SeongryongJung/llama2-7b-medmcqa-10k-asft-lora
Updated • 22
SeongryongJung/llama2-7b-medmcqa-30k-s3ft-adaptive-confidence-mix-minhard0.1-base-rkl-lora
Text Generation • Updated • 14
SeongryongJung/llama2-7b-medmcqa-30k-s3ft-adaptive-softdecay-base-rkl-beta0.3-lora
Text Generation • Updated • 18
SeongryongJung/llama2-7b-medmcqa-10k-s3ft-adaptive-softdecay-base-rkl-beta0.3-lora
Text Generation • Updated • 18
SeongryongJung/llama2-7b-medmcqa-30k-asft-lora
Text Generation • Updated • 20
SeongryongJung/llama2-7b-medmcqa-30k-lora-dft
Text Generation • Updated • 13
SeongryongJung/llama2-7b-medmcqa-30k-lora-sft
Text Generation • Updated • 15