starrylay/RC-OPD-1.7B
Text Generation • Updated • 11
RC-OPD
Qwen3-based mathematical reasoning models trained with targeted feedback on reasoning errors. LoRA adapters in 1.7B, 4B, and 8B sizes.
Note 1.7B LoRA adapter for Qwen3-1.7B. Mean Avg@4: 44.17% across AIME24, AIME25, and HMMT25.
starrylay/RC-OPD-4B
Text Generation • Updated • 10Note 4B LoRA adapter for Qwen3-4B. Mean Avg@4: 66.11% across AIME24, AIME25, and HMMT25.
starrylay/RC-OPD-8B
Text Generation • Updated • 13Note 8B LoRA adapter for Qwen3-8B. Mean Avg@4: 66.94% across AIME24, AIME25, and HMMT25.