List your company

Token-level Uncertainty Quantification on MIMIC-DI

89.11AUCROCReverse Probing (XGBoost)
41.540453.890266.2478.5898May 27, 2026

Evaluation results

MethodLinks
Reverse Probing (XGBoost)Backbone LLM=BioMistral-7B2026.0589.1154.78
Reverse Probing (CatBoost)Backbone LLM=OpenBioLLM-8B2026.0588.4850.29
Reverse Probing (XGBoost)Backbone LLM=OpenBioLLM-8B2026.0588.3349.76
Reverse Probing (CatBoost)Backbone LLM=BioMistral-7B2026.0586.9151.78
SE ProbesBackbone LLM=BioMistral-7B2026.0572.0413.13
Semantic EntropyBackbone LLM=BioMistral-7B2026.0564.488.68
SE ProbesBackbone LLM=OpenBioLLM-8B2026.0560.155.87
Decomposing UncertaintyBackbone LLM=BioMistral-7B2026.0559.388.25
Sliding Window EntropyBackbone LLM=OpenBioLLM-8B2026.0558.7712.67
Semantic EntropyBackbone LLM=OpenBioLLM-8B2026.0558.048.54
Decomposing UncertaintyBackbone LLM=OpenBioLLM-8B2026.0556.577.75
Uncertainty Aware HeadsBackbone LLM=BioMistral-7B2026.0555.088.1
Uncertainty Aware HeadsBackbone LLM=OpenBioLLM-8B2026.0554.888.52
Semantic EnergyBackbone LLM=OpenBioLLM-8B2026.0553.759.14
Attention ChainBackbone LLM=BioMistral-7B2026.0551.958.74
Sliding Window EntropyBackbone LLM=BioMistral-7B2026.0551.278.99
Attention ChainBackbone LLM=OpenBioLLM-8B2026.0549.1310.12
Token EntropyBackbone LLM=OpenBioLLM-8B2026.0548.327.67
Semantic EnergyBackbone LLM=BioMistral-7B2026.0547.148.1
Token EntropyBackbone LLM=BioMistral-7B2026.0543.376.58