List your company

Question Answering with Abstention on TriviaQA

47BASTask Calibration
14.7623.1331.539.87May 11, 2026

Evaluation results

MethodLinks
Task CalibrationModel=Gemma-3-12B, Decoding Strategy=Ours2026.0547
AncestralModel=Qwen3-30B-A3B, Decoding Strategy=Ancestral2026.0547
BeamModel=Qwen3-30B-A3B, Decoding Strategy=Beam2026.0547
AncestralModel=Gemma-3-12B, Decoding Strategy=Ancestral2026.0546
ContrastiveModel=Gemma-3-12B, Decoding Strategy=Contrastive2026.0546
ContrastiveModel=Qwen3-30B-A3B, Decoding Strategy=Contrastive2026.0546
GreedyModel=Qwen3-30B-A3B, Decoding Strategy=Greedy2026.0546
MBRModel=Qwen3-30B-A3B, Decoding Strategy=MBR2026.0546
Task CalibrationModel=Qwen3-30B-A3B, Decoding Strategy=Ours2026.0546
BeamModel=Gemma-3-12B, Decoding Strategy=Beam2026.0545
GreedyModel=Gemma-3-12B, Decoding Strategy=Greedy2026.0545
MBRModel=Gemma-3-12B, Decoding Strategy=MBR2026.0544
Task CalibrationModel=Gemma-3-4B, Decoding Strategy=Ours2026.0522
AncestralModel=Gemma-3-4B, Decoding Strategy=Ancestral2026.0521
Task CalibrationModel=Qwen-3-4B, Decoding Strategy=Ours2026.0521
BeamModel=Gemma-3-4B, Decoding Strategy=Beam2026.0520
ContrastiveModel=Gemma-3-4B, Decoding Strategy=Contrastive2026.0519
GreedyModel=Gemma-3-4B, Decoding Strategy=Greedy2026.0519
AncestralModel=Qwen-3-4B, Decoding Strategy=Ancestral2026.0519
BeamModel=Qwen-3-4B, Decoding Strategy=Beam2026.0519
ContrastiveModel=Qwen-3-4B, Decoding Strategy=Contrastive2026.0519
MBRModel=Gemma-3-4B, Decoding Strategy=MBR2026.0518
GreedyModel=Qwen-3-4B, Decoding Strategy=Greedy2026.0518
MBRModel=Qwen-3-4B, Decoding Strategy=MBR2026.0516