Planning on Sudoku
92.7AccuracyESPO
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ESPOBackbone=LLaDA-8B, Seq Len=1282025.12 | 92.7 | — | — | |
| ESPOSequence Length=128, k-shot=02026.05 | 92.7 | — | — | |
| RSPOGeneration length=2562026.05 | 92.1 | — | — | |
| GDSD w/ TLCSequence Length=256, k-shot=02026.05 | 92 | — | — | |
| GDSD w/ TLCSequence Length=Average, k-shot=02026.05 | 91 | — | — | |
| RSPOGeneration length=5122026.05 | 90.8 | — | — | |
| GDSD w/ TLCSequence Length=512, k-shot=02026.05 | 90.6 | — | — | |
| GDSD w/ TLCSequence Length=128, k-shot=02026.05 | 90.4 | — | — | |
| GDSDSequence Length=256, k-shot=02026.05 | 89.8 | — | — | |
| GDSDSequence Length=128, k-shot=02026.05 | 89.5 | — | — | |
| GDSDSequence Length=Average, k-shot=02026.05 | 89.4 | — | — | |
| FullModel=Dream-7B, Inference Strategy=Full2026.03 | 89 | — | — | |
| GDSDSequence Length=512, k-shot=02026.05 | 88.8 | — | — | |
| SPGSequence Length=256, k-shot=02026.05 | 88.2 | — | — | |
| LSPModel=Dream-7B, Inference Strategy=LSP2026.03 | 88 | 3.36 | — | |
| SPGSequence Length=128, k-shot=02026.05 | 87.3 | — | — | |
| ESPOBackbone=LLaDA-8B, Seq Len=Avg.2025.12 | 86 | — | — | |
| ESPOSequence Length=Average, k-shot=02026.05 | 86 | — | — | |
| ESPOBackbone=LLaDA-8B, Seq Len=2562025.12 | 84.7 | — | — | |
| ESPOSequence Length=256, k-shot=02026.05 | 84.7 | — | — | |
| Info-Gain SamplerK=1, Backbone=Dream-7B, Decoding Strategy=Global, Attention Type=full-attention2026.02 | 84.4 | — | — | |
| PC-SamplerK=1, Backbone=Dream-7B, Decoding Strategy=Global, Attention Type=full-attention2026.02 | 83.6 | — | — | |
| SPGSequence Length=Average, k-shot=02026.05 | 82.4 | — | — | |
| Info-Gain SamplerK=2, Backbone=Dream-7B, Decoding Strategy=Global, Attention Type=full-attention2026.02 | 82.2 | — | — | |
| KLASSK=2, Backbone=Dream-7B, Decoding Strategy=Global, Attention Type=full-attention2026.02 | 82.1 | — | — | |
| ConfidenceK=2, Backbone=Dream-7B, Decoding Strategy=Global, Attention Type=full-attention2026.02 | 81.6 | — | — | |
| ConfidenceK=1, Backbone=Dream-7B, Decoding Strategy=Global, Attention Type=full-attention2026.02 | 81.5 | — | — | |
| ESPOBackbone=LLaDA-8B, Seq Len=5122025.12 | 80.5 | — | — | |
| ESPOSequence Length=512, k-shot=02026.05 | 80.5 | — | — | |
| MarginK=1, Backbone=Dream-7B, Decoding Strategy=Global, Attention Type=full-attention2026.02 | 80.3 | — | — | |
| PC-SamplerK=2, Backbone=Dream-7B, Decoding Strategy=Global, Attention Type=full-attention2026.02 | 80.2 | — | — | |
| KLASSK=1, Backbone=Dream-7B, Decoding Strategy=Global, Attention Type=full-attention2026.02 | 79.9 | — | — | |
| EntropyK=1, Backbone=Dream-7B, Decoding Strategy=Global, Attention Type=full-attention2026.02 | 78.8 | — | — | |
| wd1Sequence Length=256, k-shot=02026.05 | 76.4 | — | — | |
| EntropyK=2, Backbone=Dream-7B, Decoding Strategy=Global, Attention Type=full-attention2026.02 | 76 | — | — | |
| MarginK=2, Backbone=Dream-7B, Decoding Strategy=Global, Attention Type=full-attention2026.02 | 74.4 | — | — | |
| ESPOBackbone=Dream-7B, Seq Len=2562025.12 | 72.3 | — | — | |
| ESPOBackbone=Dream-7B, Seq Len=Avg.2025.12 | 71.8 | — | — | |
| ESPOBackbone=Dream-7B, Seq Len=1282025.12 | 71.7 | — | — | |
| SPGSequence Length=512, k-shot=02026.05 | 71.7 | — | — | |
| ESPOBackbone=Dream-7B, Seq Len=5122025.12 | 71.3 | — | — | |
| diffu-GRPOBackbone=Dream-7B, Seq Len=2562025.12 | 69.7 | — | — | |
| diffu-GRPOBackbone=Dream-7B, Seq Len=1282025.12 | 64.4 | — | — | |
| wd1Sequence Length=512, k-shot=02026.05 | 62.8 | — | — | |
| diffu-GRPOBackbone=Dream-7B, Seq Len=Avg.2025.12 | 61.7 | — | — | |
| UniformK=1, Backbone=Dream-7B, Decoding Strategy=Global, Attention Type=full-attention2026.02 | 60.4 | — | — | |
| wd1Sequence Length=Average, k-shot=02026.05 | 53.9 | — | — | |
| UniformK=2, Backbone=Dream-7B, Decoding Strategy=Global, Attention Type=full-attention2026.02 | 52.8 | — | — | |
| diffu-GRPOBackbone=Dream-7B, Seq Len=5122025.12 | 51.1 | — | — | |
| wd1Backbone=Dream-7B, Seq Len=2562025.12 | 39.2 | — | — | |
| LSPModel=LLaDA-8B, Inference Strategy=LSP2026.03 | 36 | 3 | — | |
| FullModel=LLaDA-8B, Inference Strategy=Full2026.03 | 35 | — | — | |
| PC-Sampler w/ E-BONModel=LLaDA-1.5-8B, Decoding Strategy=PC-Sampler + E-BON2025.12 | 34.2 | — | — | |
| MarginModel=LLaDA-1.5-8B, Decoding Strategy=Margin2025.12 | 33.6 | — | — | |
| PC-Sampler w/ E-SMCModel=LLaDA-1.5-8B, Decoding Strategy=PC-Sampler + E-SMC2025.12 | 33.6 | — | — | |
| PC-SamplerModel=LLaDA-1.5-8B, Decoding Strategy=PC-Sampler2025.12 | 33.4 | — | — | |
| wd1Backbone=Dream-7B, Seq Len=Avg.2025.12 | 33 | — | — | |
| DMPOSequence Length=128, k-shot=02026.05 | 32.8 | — | — | |
| wd1Backbone=Dream-7B, Seq Len=5122025.12 | 30.3 | — | — | |
| wd1Backbone=Dream-7B, Seq Len=1282025.12 | 29.5 | — | — | |
| PC-Sampler w/ E-SMCModel=LLaDA-Instruct-8B, Decoding Strategy=PC-Sampler + E-SMC2025.12 | 29.2 | — | — | |
| PC-Sampler w/ E-BONModel=LLaDA-Instruct-8B, Decoding Strategy=PC-Sampler + E-BON2025.12 | 28.2 | — | — | |
| P2Model=LLaDA-1.5-8B, Decoding Strategy=P22025.12 | 27.8 | — | — | |
| PC-SamplerModel=LLaDA-Instruct-8B, Decoding Strategy=PC-Sampler2025.12 | 27.6 | — | — | |
| BGPOBase Model=LLaDA-8B-Instruct, RL Algorithm=BGPO2025.10 | 26.9 | — | — | |
| diffu-GRPOBackbone=LLaDA-8B, Seq Len=1282025.12 | 26.7 | — | — | |
| diffu-GRPO (d1)Sequence Length=128, k-shot=02026.05 | 26.7 | — | — | |
| diffu-GRPOBase Model=LLaDA-8B-Instruct, RL Algorithm=diffu-GRPO2025.10 | 26.7 | — | — | |
| MarginModel=LLaDA-Instruct-8B, Decoding Strategy=Margin2025.12 | 26.6 | — | — | |
| VRPO-OLBase Model=LLaDA-8B-Instruct, RL Algorithm=VRPO-OL, Sampling size (nt)=42025.10 | 26.1 | — | — | |
| BGPO (off-policy)Base Model=LLaDA-8B-Instruct, RL Algorithm=BGPO, Mode=off-policy2025.10 | 26 | — | — | |
| TraceRLGeneration length=2562026.05 | 25.6 | — | — | |
| DMPOSequence Length=Average, k-shot=02026.05 | 25.6 | — | — | |
| TraceRLGeneration length=5122026.05 | 25.4 | — | — | |
| wd12025.10 | 25.2 | — | — | |
| ConfidenceModel=LLaDA-1.5-8B, Decoding Strategy=Confidence2025.12 | 24.8 | — | — | |
| LLaDABackbone=LLaDA-8B, Seq Len=1282025.12 | 24.8 | — | — | |
| LLaDA-8BSequence Length=128, k-shot=02026.05 | 24.8 | — | — | |
| wd1Backbone=LLaDA-8B, Seq Len=5122025.12 | 24.6 | — | — | |
| wd1Generation length=5122026.05 | 24.6 | — | — | |
| DMPOSequence Length=256, k-shot=02026.05 | 24.6 | — | — | |
| EB-SamplerModel=LLaDA-Instruct-8B, Decoding Strategy=EB-Sampler2025.12 | 24.2 | — | — | |
| diffu-GRPOBackbone=LLaDA-8B, Seq Len=2562025.12 | 24.1 | — | — | |
| d1Generation length=2562026.05 | 24.1 | — | — | |
| diffu-GRPO (d1)Sequence Length=256, k-shot=02026.05 | 24.1 | — | — | |
| P2Model=LLaDA-Instruct-8B, Decoding Strategy=P22025.12 | 24 | — | — | |
| d-OPSDSeq Len=1282026.06 | 23.9 | — | — | |
| ConfidenceModel=LLaDA-Instruct-8B, Decoding Strategy=Confidence2025.12 | 23.8 | — | — | |
| wd1Backbone=LLaDA-8B, Seq Len=Avg.2025.12 | 23.1 | — | — | |
| wd1Backbone=LLaDA-8B, Seq Len=1282025.12 | 22.6 | — | — | |
| wd1Sequence Length=128, k-shot=02026.05 | 22.6 | — | — | |
| diffu-GRPOBackbone=LLaDA-8B, Seq Len=Avg.2025.12 | 22.2 | — | — | |
| diffu-GRPO (d1)Sequence Length=Average, k-shot=02026.05 | 22.2 | — | — | |
| wd1Backbone=LLaDA-8B, Seq Len=2562025.12 | 22 | — | — | |
| wd1Generation length=2562026.05 | 22 | — | — | |
| d-OPSDSeq Len=2562026.06 | 20.6 | — | — | |
| SAPOGeneration length=2562026.05 | 20.3 | — | — | |
| DMPOSequence Length=512, k-shot=02026.05 | 20 | — | — | |
| diffu-GRPOSeq Len=1282026.06 | 18.4 | — | — | |
| d1-LLaDA2025.10 | 16.7 | — | — |