List your company

Reasoning on GPQA Diamond (Acc, #Tok, ∆%)

58.8AccuracyVanilla
10.4422.99535.5548.105Jun 2, 2026

Evaluation results

MethodLinks
VanillaReasoner=Qwen3-8B2026.0658.89,654—
ACTSReasoner=Qwen3-8B2026.0658.66,61531.5
ACTSπSFTReasoner=Qwen3-8B2026.0658.47,76119.6
CoDReasoner=Qwen3-8B2026.0658.35,78940
DEERReasoner=Qwen3-8B2026.0657.17,67120.5
NoThinkReasoner=Qwen3-8B2026.0648.11,62883.1
BudgetGuidanceReasoner=Qwen3-8B2026.0647.82,30876.1
ACTSReasoner=DeepSeek-R1-Distill-Qwen-7B2026.0646.84,40447.7
ACTSπSFTReasoner=DeepSeek-R1-Distill-Qwen-7B2026.0644.44,65644.7
VanillaReasoner=DeepSeek-R1-Distill-Qwen-7B2026.0638.98,422—
CoDReasoner=DeepSeek-R1-Distill-Qwen-7B2026.0636.75,40235.9
BudgetGuidanceReasoner=DeepSeek-R1-Distill-Qwen-7B2026.0636.54,56545.8
NoThinkReasoner=DeepSeek-R1-Distill-Qwen-7B2026.0635.470091.7
DEERReasoner=DeepSeek-R1-Distill-Qwen-7B2026.0634.93,82554.6
ACTSπSFTReasoner=DeepSeek-R1-Distill-Qwen-1.5B2026.0630.13,75859.5
ACTSReasoner=DeepSeek-R1-Distill-Qwen-1.5B2026.0630.13,76659.4
CoDReasoner=DeepSeek-R1-Distill-Qwen-1.5B2026.0620.97,99313.9
NoThinkReasoner=DeepSeek-R1-Distill-Qwen-1.5B2026.0619.297789.5
VanillaReasoner=DeepSeek-R1-Distill-Qwen-1.5B2026.0618.29,285—
DEERReasoner=DeepSeek-R1-Distill-Qwen-1.5B2026.0612.34,76948.6