List your company

Token-level Uncertainty Quantification on Generated-DI

93.64AUCROCReverse Probing (XGBoost)
38.488852.806967.12581.4431May 27, 2026

Evaluation results

MethodLinks
Reverse Probing (XGBoost)Backbone LLM=OpenBioLLM-8B2026.0593.6428.92
Reverse Probing (CatBoost)Backbone LLM=OpenBioLLM-8B2026.0591.4428.95
Reverse Probing (XGBoost)Backbone LLM=BioMistral-7B2026.0588.1923.16
Reverse Probing (CatBoost)Backbone LLM=BioMistral-7B2026.0584.3528.05
Sliding Window EntropyBackbone LLM=OpenBioLLM-8B2026.0569.410.41
Uncertainty Aware HeadsBackbone LLM=OpenBioLLM-8B2026.0564.036.82
Uncertainty Aware HeadsBackbone LLM=BioMistral-7B2026.0561.55.75
Decomposing UncertaintyBackbone LLM=OpenBioLLM-8B2026.0559.5914.1
Decomposing UncertaintyBackbone LLM=BioMistral-7B2026.0559.2513.22
SE ProbesBackbone LLM=OpenBioLLM-8B2026.0559.0915.53
Sliding Window EntropyBackbone LLM=BioMistral-7B2026.0558.846.78
Semantic EnergyBackbone LLM=OpenBioLLM-8B2026.0555.344.44
Token EntropyBackbone LLM=OpenBioLLM-8B2026.0553.634.4
Semantic EntropyBackbone LLM=BioMistral-7B2026.0553.6110.5
Semantic EntropyBackbone LLM=OpenBioLLM-8B2026.0553.6110.5
Attention ChainBackbone LLM=OpenBioLLM-8B2026.0550.984.3
Semantic EnergyBackbone LLM=BioMistral-7B2026.0546.974.36
Attention ChainBackbone LLM=BioMistral-7B2026.0545.493.88
Token EntropyBackbone LLM=BioMistral-7B2026.0544.213.66
SE ProbesBackbone LLM=BioMistral-7B2026.0540.6110.75