List your company

Question Answering on ARC (pass@1)

88.68pass@1MCDIFFUSE
5.594427.164748.73570.3053Jan 12, 2026Jan 17, 2026Jan 22, 2026Jan 28, 2026Feb 2, 2026Feb 7, 2026Feb 13, 2026

Evaluation results

MethodLinks
MCDIFFUSEModel Type=Diffusion, Parameters=7B, Slot Ordering Strategy=MCTS2026.0288.68
Qwen3 8BModel Type=Autoregressive, Parameters=8B2026.0288.4
ReFusion 7BModel Type=Diffusion, Parameters=7B, Slot Ordering Strategy=Vanilla2026.0287.98
Qwen2.5 7BModel Type=Autoregressive, Parameters=7B2026.0287.8
ReFusion 7BModel Type=Diffusion, Parameters=7B, Slot Ordering Strategy=Sequential2026.0287.09
SCR (Ours)Backbone=Qwen2.5-7B-Instruct2026.0186.43
GRPOBackbone=Qwen2.5-7B-Instruct2026.0185.49
Dream 7BModel Type=Diffusion, Parameters=7B2026.0284.95
SFT + GRPOBackbone=Llama3.1-8B-Instruct2026.0184.73
SCR (Ours)Backbone=Llama3.1-8B-Instruct2026.0184.56
BaseBackbone=Qwen2.5-7B-Instruct2026.0184.39
LLaDA 8BModel Type=Diffusion, Parameters=8B2026.0284.09
Llama-3.1 8BModel Type=Autoregressive, Parameters=8B2026.0283.32
SCR-Stage IBackbone=Llama3.1-8B-Instruct2026.0182.17
GRPOBackbone=Qwen2.5-3B-Instruct2026.0181.14
ReFusion 7BModel Type=Diffusion, Parameters=7B, Slot Ordering Strategy=Random2026.0280.98
SFT + GRPOBackbone=Qwen2.5-7B-Instruct2026.0180.63
SCR-Stage IBackbone=Qwen2.5-7B-Instruct2026.0180.63
SCR (Ours)Backbone=Qwen2.5-3B-Instruct2026.0179.35
SFT + GRPOBackbone=Qwen2.5-3B-Instruct2026.0179.01
SCR-SFTBackbone=Llama3.1-8B-Instruct2026.0174.06
SCR-SFTBackbone=Qwen2.5-7B-Instruct2026.0172.95
GRPOBackbone=Llama3.1-8B-Instruct2026.0170.31
BaseBackbone=Qwen2.5-3B-Instruct2026.0169.2
SCR-Stage IBackbone=Qwen2.5-3B-Instruct2026.0167.32
Self-RefineBackbone=Qwen2.5-3B-Instruct2026.0163.57
Self-RefineBackbone=Qwen2.5-7B-Instruct2026.0157.51
BaseBackbone=Llama3.1-8B-Instruct2026.0157.08
SCR-SFTBackbone=Qwen2.5-3B-Instruct2026.0129.69
Self-RefineBackbone=Llama3.1-8B-Instruct2026.018.79