List your company

Reasoning on GPQA Diamond (Pass@1, FS, N/T, D/T, Repro)

66.8Pass@1 AccuracyRED
58.58460.71762.8564.983Apr 3, 2026

Evaluation results

MethodLinks
REDModel=Qwen3-32B-Thinking2026.0466.82.33.22.32
S-GRROModel=Qwen3-32B-Thinking2026.0466.54.65.33.85.6
REDModel=R1-Llama-70B2026.0466.22.93.52.83
RL + LPModel=Qwen3-32B-Thinking2026.04664.45.23.55.3
VanillaModel=Qwen3-32B-Thinking2026.0465.35.15.646.1
S-GRROModel=R1-Llama-70B2026.0465.25.95.74.68.6
RL + LPModel=R1-Llama-70B2026.04655.75.64.48.3
DEERModel=Qwen3-32B-Thinking2026.0464.84.75.44.16.2
VanillaModel=R1-Llama-70B2026.0464.56.56.24.89.4
DEERModel=R1-Llama-70B2026.0463.15.95.94.99.6
REDModel=R1-Qwen-32B2026.0461.82.43.52.73
S-GRROModel=R1-Qwen-32B2026.0461.34.75.64.26.6
VanillaModel=R1-Qwen-32B2026.0460.85.35.94.47.1
RL + LPModel=R1-Qwen-32B2026.0460.44.65.53.96.4
DEERModel=R1-Qwen-32B2026.0458.94.95.74.47.3