Knowledge Reasoning on GPQA Diamond
75.4Accuracy (avg@8)TRICE-30B
Evaluation Results
| Method | Links | |
|---|---|---|
| TRICE-30BUse Tool=true2026.05 | 75.4 | |
| Qwen3-30B-A3B-Thinking-2507Use Tool=false2026.05 | 71.2 | |
| TRICE-4BUse Tool=true2026.05 | 68.8 | |
| Qwen3-4B-Thinking-2507Use Tool=false2026.05 | 64.4 | |
| Qwen3Params=32B2026.04 | 64.1 | |
| I-DLMParams=32B, Decoding=ISD (N=4, sampling)2026.04 | 62.1 | |
| Qwen3Params=8B2026.04 | 58.9 | |
| I-DLMParams=8B, Decoding=ISD (N=4, sampling)2026.04 | 55.6 | |
| DeepSeek-R1-Distill-Qwen-7BArchitecture=Dense, # Activated Params=7B, # Total Params=7B2025.12 | 47.1 | |
| Sigma-MoE-TinyArchitecture=MoE, # Activated Params=0.5B, # Total Params=20B2025.12 | 46.4 | |
| LLaDA-2.1-miniParams=16B2026.04 | 46 | |
| DeepSeek-R1-Distill-Llama-8BArchitecture=Dense, # Activated Params=8B, # Total Params=8B2025.12 | 43.2 | |
| SDARParams=8B2026.04 | 40.2 | |
| Qwen3-1.7BArchitecture=Dense, # Activated Params=1.7B, # Total Params=1.7B2025.12 | 40.1 | |
| Phi-3.5-MoEArchitecture=MoE, # Activated Params=6.6B, # Total Params=42B2025.12 | 36.8 | |
| SDARParams=30B2026.04 | 36.7 |