List your company

Scientific Reasoning on PIQA (test)

90.8AccuracyRCSD
89.13689.5689090.432Jun 17, 2026

Evaluation results

MethodLinks
RCSDBackbone=Qwen3-8B, Training Steps=1002026.0690.8
GRPOBackbone=Qwen3-8B, Training Steps=5002026.0690.3
Rubric-GRPOBackbone=Qwen3-8B2026.0690.3
Qwen3-8BBackbone=Qwen3-8B2026.0690.2
OPSDBackbone=Qwen3-8B, Training Steps=1002026.0690.1
SFTBackbone=Qwen3-8B2026.0689.2