List your company

Ordinal Regression on STSB

0.65L1 LossTask Calibration
0.62240.80870.9951.1813May 11, 2026

Evaluation results

MethodLinks
Task CalibrationModel=Gemma-3-12B, Decoding Strategy=Ours2026.050.65
Task CalibrationModel=Qwen3-30B-A3B, Decoding Strategy=Ours2026.050.67
Task CalibrationModel=Qwen-3-4B, Decoding Strategy=Ours2026.050.68
Task CalibrationModel=Gemma-3-4B, Decoding Strategy=Ours2026.050.8
BeamModel=Gemma-3-12B, Decoding Strategy=Beam2026.050.98
ContrastiveModel=Gemma-3-12B, Decoding Strategy=Contrastive2026.050.99
GreedyModel=Gemma-3-12B, Decoding Strategy=Greedy2026.051
AncestralModel=Gemma-3-12B, Decoding Strategy=Ancestral2026.051.01
MBRModel=Gemma-3-12B, Decoding Strategy=MBR2026.051.01
AncestralModel=Gemma-3-4B, Decoding Strategy=Ancestral2026.051.05
BeamModel=Gemma-3-4B, Decoding Strategy=Beam2026.051.06
GreedyModel=Gemma-3-4B, Decoding Strategy=Greedy2026.051.07
MBRModel=Gemma-3-4B, Decoding Strategy=MBR2026.051.07
ContrastiveModel=Gemma-3-4B, Decoding Strategy=Contrastive2026.051.08
ContrastiveModel=Qwen3-30B-A3B, Decoding Strategy=Contrastive2026.051.17
MBRModel=Qwen3-30B-A3B, Decoding Strategy=MBR2026.051.18
AncestralModel=Qwen3-30B-A3B, Decoding Strategy=Ancestral2026.051.19
BeamModel=Qwen3-30B-A3B, Decoding Strategy=Beam2026.051.19
GreedyModel=Qwen3-30B-A3B, Decoding Strategy=Greedy2026.051.19
AncestralModel=Qwen-3-4B, Decoding Strategy=Ancestral2026.051.33
BeamModel=Qwen-3-4B, Decoding Strategy=Beam2026.051.33
ContrastiveModel=Qwen-3-4B, Decoding Strategy=Contrastive2026.051.33
MBRModel=Qwen-3-4B, Decoding Strategy=MBR2026.051.33
GreedyModel=Qwen-3-4B, Decoding Strategy=Greedy2026.051.34