List your company

Question Answering on OpenBookQA (Acc, ECE, Brier)

0.874AccuracyCoRA
0.784560.807780.8310.85422Jun 12, 2026

Evaluation results

MethodLinks
CoRABackbone Model=Ministral 3 8B2026.060.8740.11720.2396
SFTBackbone Model=Ministral 3 8B2026.060.8720.11540.2423
GRPOBackbone Model=Ministral 3 8B2026.060.8640.11540.2478
CoRABackbone Model=Qwen2.5 7B2026.060.860.11890.2557
GRPOBackbone Model=Gemma 2 9B Instruct2026.060.8420.14520.3023
CoRABackbone Model=Gemma 2 9B Instruct2026.060.840.14140.3021
BaseBackbone Model=Ministral 3 8B2026.060.8340.13940.305
SFTBackbone Model=Qwen2.5 7B2026.060.8280.12710.2974
GRPOBackbone Model=Qwen2.5 7B2026.060.8260.10730.278
BaseBackbone Model=Qwen2.5 7B2026.060.8080.10650.3045
SFTBackbone Model=Gemma 2 9B Instruct2026.060.8060.19050.3786
BaseBackbone Model=Gemma 2 9B Instruct2026.060.7880.19160.3877