Sft-new ssurface/qwen3-4b-sft-l5-merged 4B • Updated 17 days ago • 21 ssurface/qwen3-4b-sft-l3-merged 4B • Updated 17 days ago • 20 ssurface/qwen3-4b-sft-l4-merged 4B • Updated 17 days ago • 18 ssurface/qwen3-4b-sft-l2-merged 4B • Updated 17 days ago • 19
GRPO no length punshiment SFT GSM8K ssurface/qwen3-4b-grpo-nolength-l1 Text Generation • 4B • Updated 24 days ago • 16 ssurface/qwen3-4b-grpo-nolength-l2 Text Generation • 4B • Updated 24 days ago • 23 ssurface/qwen3-4b-grpo-nolength-l3 Text Generation • 4B • Updated 24 days ago • 22 ssurface/qwen3-4b-grpo-nolength-l4 Text Generation • 4B • Updated 24 days ago • 20
GRPO SFT GSM8K ssurface/qwen3-4b-grpo-l1 Text Generation • 4B • Updated 30 days ago • 21 ssurface/qwen3-4b-grpo-l2 Text Generation • 4B • Updated 30 days ago • 21 ssurface/qwen3-4b-grpo-l3 Text Generation • 4B • Updated 30 days ago • 26 ssurface/qwen3-4b-grpo-l4 Text Generation • 4B • Updated 30 days ago • 19
GRPO SFT-Length-Punishment-GDPO SFT GSM8K ssurface/qwen3-4b-gdpo-length-sft-l1 Text Generation • 4B • Updated 23 days ago • 282 ssurface/qwen3-4b-gdpo-length-sft-l2 Text Generation • 4B • Updated 23 days ago • 288 ssurface/qwen3-4b-gdpo-length-sft-l3 Text Generation • 4B • Updated 23 days ago • 287 ssurface/qwen3-4b-gdpo-length-sft-l4 Text Generation • 4B • Updated 23 days ago • 300
GRPO abstract reward SFT GSM8K ssurface/qwen3-4b-grpo-abstract-l1 Text Generation • 4B • Updated 26 days ago • 25 ssurface/qwen3-4b-grpo-abstract-l2 Text Generation • 4B • Updated 26 days ago • 24 ssurface/qwen3-4b-grpo-abstract-l3 Text Generation • 4B • Updated 26 days ago • 25 ssurface/qwen3-4b-grpo-abstract-l4 Text Generation • 4B • Updated 26 days ago • 25
Qwen3-4B CoT Compression Study LoRA adapters trained for 5 progressively shorter chain-of-thought styles on GSM8K, plus the eval artifacts behind the Pareto curve. ssurface/qwen3-4b-cot-compress-l1 Text Generation • Updated May 25 • 2 ssurface/qwen3-4b-cot-compress-l2 Text Generation • Updated May 25 • 1 ssurface/qwen3-4b-cot-compress-l3 Text Generation • Updated May 25 • 1 ssurface/qwen3-4b-cot-compress-l4 Text Generation • Updated May 25 • 1
Sft-new ssurface/qwen3-4b-sft-l5-merged 4B • Updated 17 days ago • 21 ssurface/qwen3-4b-sft-l3-merged 4B • Updated 17 days ago • 20 ssurface/qwen3-4b-sft-l4-merged 4B • Updated 17 days ago • 18 ssurface/qwen3-4b-sft-l2-merged 4B • Updated 17 days ago • 19
GRPO SFT-Length-Punishment-GDPO SFT GSM8K ssurface/qwen3-4b-gdpo-length-sft-l1 Text Generation • 4B • Updated 23 days ago • 282 ssurface/qwen3-4b-gdpo-length-sft-l2 Text Generation • 4B • Updated 23 days ago • 288 ssurface/qwen3-4b-gdpo-length-sft-l3 Text Generation • 4B • Updated 23 days ago • 287 ssurface/qwen3-4b-gdpo-length-sft-l4 Text Generation • 4B • Updated 23 days ago • 300
GRPO no length punshiment SFT GSM8K ssurface/qwen3-4b-grpo-nolength-l1 Text Generation • 4B • Updated 24 days ago • 16 ssurface/qwen3-4b-grpo-nolength-l2 Text Generation • 4B • Updated 24 days ago • 23 ssurface/qwen3-4b-grpo-nolength-l3 Text Generation • 4B • Updated 24 days ago • 22 ssurface/qwen3-4b-grpo-nolength-l4 Text Generation • 4B • Updated 24 days ago • 20
GRPO abstract reward SFT GSM8K ssurface/qwen3-4b-grpo-abstract-l1 Text Generation • 4B • Updated 26 days ago • 25 ssurface/qwen3-4b-grpo-abstract-l2 Text Generation • 4B • Updated 26 days ago • 24 ssurface/qwen3-4b-grpo-abstract-l3 Text Generation • 4B • Updated 26 days ago • 25 ssurface/qwen3-4b-grpo-abstract-l4 Text Generation • 4B • Updated 26 days ago • 25
GRPO SFT GSM8K ssurface/qwen3-4b-grpo-l1 Text Generation • 4B • Updated 30 days ago • 21 ssurface/qwen3-4b-grpo-l2 Text Generation • 4B • Updated 30 days ago • 21 ssurface/qwen3-4b-grpo-l3 Text Generation • 4B • Updated 30 days ago • 26 ssurface/qwen3-4b-grpo-l4 Text Generation • 4B • Updated 30 days ago • 19
Qwen3-4B CoT Compression Study LoRA adapters trained for 5 progressively shorter chain-of-thought styles on GSM8K, plus the eval artifacts behind the Pareto curve. ssurface/qwen3-4b-cot-compress-l1 Text Generation • Updated May 25 • 2 ssurface/qwen3-4b-cot-compress-l2 Text Generation • Updated May 25 • 1 ssurface/qwen3-4b-cot-compress-l3 Text Generation • Updated May 25 • 1 ssurface/qwen3-4b-cot-compress-l4 Text Generation • Updated May 25 • 1
