List your company

Question Answering on MIP and AbstentionBench

100Answer ScoreBase
92.09694.14896.298.252May 27, 2026

Evaluation results

MethodLinks
BaseBackbone=DeepSeek-R1-Distill-Qwen-14B2026.0510091.81,687.30.544
BaseBackbone=Qwen3-30B-A3B-Thinking2026.0510097.52,658.50.367
PromptingBackbone=DeepSeek-R1-Distill-Qwen-14B2026.0599.689.11,046.50.851
Plain RLBackbone=DeepSeek-R1-Distill-Qwen-14B2026.0599.490.31,769.70.51
Plain RLBackbone=Qwen3-30B-A3B-Thinking2026.059996.11,694.10.567
PromptingBackbone=Qwen3-30B-A3B-Thinking2026.0598.8941,440.10.653
JTSBackbone=Qwen3-30B-A3B-Thinking2026.0594.793.4810.41.152
JTSBackbone=DeepSeek-R1-Distill-Qwen-14B2026.0592.486.8844.71.028