List your company

Graduate-level Science QA on GPQA (Accuracy)

52.02AccuracyParallel-Synthesis
2.703215.506628.3141.1134May 14, 2026May 18, 2026May 23, 2026May 28, 2026Jun 2, 2026Jun 7, 2026Jun 12, 2026

Evaluation results

MethodLinks
Parallel-SynthesisBackbone=Qwen3-14B, Workers=3, Merging=Weighted averaging (lambda=0.5)2026.0652.02
Text-SerializationBackbone=Qwen3-14B, Workers=32026.0650
VotingBackbone=Qwen3-14B, Workers=32026.0648.99
KVLINKBackbone=Qwen3-14B, Workers=32026.0635.35
SingleBackbone=Qwen3-14B, Workers=12026.0631.31
Qwen2.5-3B-InstructBackbone=Qwen2.5-3B, Training Protocol=Instruct2026.0530.8
Qwen2.5-3B-RLVRBackbone=Qwen2.5-3B, Training Protocol=RLVR2026.0529.3
Qwen2.5-3B-GRLOBackbone=Qwen2.5-3B, Training Protocol=GRLO2026.0529.3
Qwen2.5-3B-GRLO+RLVRBackbone=Qwen2.5-3B, Training Protocol=GRLO+RLVR2026.0527.8
Qwen2.5-3B-BaseBackbone=Qwen2.5-3B, Training Protocol=Base2026.0522.2
CacheBlendBackbone=Qwen3-14B, Workers=32026.0612.63
APEBackbone=Qwen3-14B, Workers=32026.0610.1
Qwen2.5-3B-MathSFTBackbone=Qwen2.5-3B, Training Protocol=MathSFT2026.054.6