List your company
ResearchFormerly WizWand

Science Question Answering on GPQA Diamond (accuracy)

77AccuracyDeepseek-V3.2
17.7233.1148.563.89Nov 25, 2025Dec 26, 2025Jan 27, 2026Feb 28, 2026Apr 1, 2026May 3, 2026Jun 4, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Deepseek-V3.2Base Model=Deepseek-V3.22026.0577
Kimi-K2.5Base Model=Kimi-K2.52026.0576.4
Qwen3.5-9B + MaRBase Model=Qwen3.5-9B, Optimization Method=MaR2026.0574.7
Qwen3.5-122BBase Model=Qwen3.5-122B2026.0573.7
Qwen3.5-397BBase Model=Qwen3.5-397B2026.0573.3
Qwen3.5-35BBase Model=Qwen3.5-35B2026.0571.2
Qwen3.5-9B + DAPOBase Model=Qwen3.5-9B, Optimization Method=DAPO2026.0571.1
Qwen3.5-4B + MaRBase Model=Qwen3.5-4B, Optimization Method=MaR2026.0568.7
GLM-5.1Base Model=GLM-5.12026.0567.5
Qwen3.5-9BBase Model=Qwen3.5-9B2026.0567
Qwen3.5-4B + DAPOBase Model=Qwen3.5-4B, Optimization Method=DAPO2026.0565.9
GPT-OSS-120BBase Model=GPT-OSS-120B2026.0565.7
Qwen3.5-4BBase Model=Qwen3.5-4B2026.0563.3
ReLATModel=Qwen3-14B2026.0656.5
LatentMASModel=Qwen3-14B2026.0652
TextMASModel=Qwen3-14B2026.0651.5
SingleModel=Qwen3-14B2026.0648.5
ReLATModel=Qwen3-8B2026.0648
LatentMASModel=Qwen3-8B2026.0645.5
TextMASModel=Qwen3-8B2026.0643.4
ReLATModel=Qwen3-4B2026.0642.4
LatentMASModel=Qwen3-4B2026.0641.9
TextMASModel=Qwen3-4B2026.0640.4
SingleModel=Qwen3-8B2026.0639.9
ReLATModel=DeepSeek-R1-Distill-Qwen-7B2026.0637.9
SingleModel=Qwen3-4B2026.0636.3
TextMASModel=DeepSeek-R1-Distill-Qwen-7B2026.0635.3
LatentMASModel=DeepSeek-R1-Distill-Qwen-7B2026.0629.3
evaluation-instructed prompt optimizationBackbone=LLaMA-3, Optimization Type=Query-dependent2025.1129
Pro-RefineBackbone=LLaMA-3.1, Optimization Type=Query-dependent2025.1129
LLM onlyBackbone=LLaMA-3, Optimization Type=None2025.1128
Pro-RefineBackbone=LLaMA-3, Optimization Type=Query-dependent2025.1128
SingleModel=DeepSeek-R1-Distill-Qwen-7B2026.0627.7
APEBackbone=LLaMA-3, Optimization Type=Template2025.1127
evaluation-instructed prompt optimizationBackbone=LLaMA-3.1, Optimization Type=Query-dependent2025.1127
TextGradBackbone=LLaMA-3.1, Optimization Type=Template2025.1127
Self-RefineBackbone=LLaMA-3, Optimization Type=Query-dependent2025.1126
TextGradBackbone=LLaMA-3, Optimization Type=Template2025.1126
LLM onlyBackbone=LLaMA-3.1, Optimization Type=None2025.1126
Self-RefineBackbone=LLaMA-3.1, Optimization Type=Query-dependent2025.1126
APEBackbone=LLaMA-3.1, Optimization Type=Template2025.1124
evaluation-instructed prompt optimizationBackbone=GPT-4o, Optimization Type=Query-dependent2025.1124
TextGradBackbone=GPT-4o, Optimization Type=Template2025.1123
LLM onlyBackbone=GPT-4o, Optimization Type=None2025.1122
APEBackbone=GPT-4o, Optimization Type=Template2025.1122
Pro-RefineBackbone=GPT-4o, Optimization Type=Query-dependent2025.1121
Self-RefineBackbone=GPT-4o, Optimization Type=Query-dependent2025.1120