List your company

Scientific Reasoning on SCIBENCH (test)

70.8AccuracyRCSD
65.18466.64268.169.558Jun 17, 2026

Evaluation results

MethodLinks
RCSDBackbone=Qwen3-8B, Training Steps=1002026.0670.8
Rubric-GRPOBackbone=Qwen3-8B2026.0670.2
GRPOBackbone=Qwen3-8B, Training Steps=5002026.0669.7
Qwen3-8BBackbone=Qwen3-8B2026.0669.1
OPSDBackbone=Qwen3-8B, Training Steps=1002026.0668.7
SFTBackbone=Qwen3-8B2026.0665.4