jhyuckkim/Qwen3-30B-A3B-Dense-3B-DO-ACP-K8
Text Generation • 3B • Updated • 380
MoE-to-Dense: pruned and distilled dense students
Dense students pruned and distilled from Qwen3-30B-A3B, DeepSeek-V2-Lite and GPT-OSS-20B. Research artifacts for arXiv:2605.28207.
Note paper headline result (Table 6)
jhyuckkim/DeepSeek-V2-Lite-Dense-3B-DO-ACP-K6
Text Generation • 3B • Updated • 341Note best DeepSeek student (Table 17)
jhyuckkim/GPT-OSS-20B-Dense-4B-DO-ACP-K4
Text Generation • 4B • Updated • 384 • 1Note best GPT-OSS student (Table 18)
-
jhyuckkim/Qwen3-30B-A3B-Dense-3B-SF-K16
Text Generation • 3B • Updated • 368 -
jhyuckkim/Qwen3-32B-Dense-3B-D2D
Text Generation • 3B • Updated • 509 -
jhyuckkim/DeepSeek-V2-Lite-Dense-3B-ACP-K6
Text Generation • 3B • Updated • 345 -
jhyuckkim/DeepSeek-V2-Lite-Dense-3B-CP-K6
Text Generation • 3B • Updated • 508 -
jhyuckkim/DeepSeek-V2-Lite-Dense-3B-SF-K6
Text Generation • 3B • Updated • 500 -
jhyuckkim/GPT-OSS-20B-Dense-4B-ACP-K4
Text Generation • 4B • Updated • 385 -
jhyuckkim/GPT-OSS-20B-Dense-4B-CP-K4
Text Generation • 4B • Updated • 496 -
jhyuckkim/GPT-OSS-20B-Dense-4B-SF-K4
Text Generation • 4B • Updated • 510