List your company

Uncertainty Estimation on emrQA

77.5AUROCLogit Magnitude
35.69246.54657.468.254May 7, 2026

Evaluation results

MethodLinks
Logit MagnitudeLLM=Gemma4-4B, generation=Full (τ = T)2026.0577.576.471.121.5
Logit MagnitudeLLM=Qwen3.5-4B, generation=Full (τ = T)2026.0577.560.571.343.3
Self-CertaintyLLM=Qwen3.5-4B2026.0577.160.370.743.3
Logit MagnitudeLLM=Qwen3.5-4B, generation=Early-stopped2026.057760.471.138.2
Self-CertaintyLLM=Gemma4-4B2026.0576.875.570.821.5
Sequence EntropyLLM=Qwen3.5-4B2026.0576.56069.843.3
Logit MagnitudeLLM=Gemma4, Size=31B2026.0576.17170.319.5
Self-CertaintyLLM=Gemma4, Size=31B2026.0575.370.571.419.5
Sequence EntropyLLM=Gemma4-4B2026.0574.574.568.921.5
Self-CertaintyBase Model=Llama3, LLM Size=8B2026.0574.260.268.120.4
Logit MagnitudeLLM=Gemma4-4B, generation=Early-stopped2026.05747568.717.8
MetaUELLM=Gemma4-4B2026.0573.358.866.70
MetaUELLM=Qwen3.5-4B2026.0573.358.866.70
Logit MagnitudeLLM=Qwen3.5, Size=9B2026.0573.364.168.341.1
Logit MagnitudeLLM=Gemma4, Size=26B*2026.0573.172.267.521.5
Self-CertaintyLLM=Gemma4, Size=26B*2026.0572.971.468.721.5
Self-CertaintyLLM=Qwen3.5, Size=9B2026.0572.963.86841.1
Self-CertaintyBase Model=Llama3, LLM Size=3B2026.0572.56066.627.4
Logit MagnitudeLLM=Qwen3.5, Size=35B*2026.057255.468.738
Hidden-State ScoreLLM=Qwen3.5-4B2026.0571.956.967.143.3
LogTokULLM=Qwen3.5-4B2026.0571.856.56743.3
MetaUELLM=Gemma4, Size=31B2026.0571.768.566.20
Self-CertaintyLLM=Qwen3.5, Size=35B*2026.0571.555.868.238
MetaUELLM=Qwen3.5, Size=9B2026.057164.264.80
MetaUELLM=Qwen3.5, Size=35B*2026.0570.255.164.60
MetaUELLM=Gemma4, Size=26B*2026.0569.670.764.80
LogTokULLM=Qwen3.5, Size=9B2026.0569.661.266.441.1
LogTokULLM=Gemma4, Size=26B*2026.0567.267.165.821.5
Semantic EntropyLLM=Gemma4-4B2026.0567.170.862.7216.1
Semantic EntropyLLM=Qwen3.5-4B2026.0566.754.761.8430.2
Logit MagnitudeBase Model=Llama3, LLM Size=3B2026.0565.555.261.827.4
Self-ConsistencyLLM=Gemma4-4B2026.0565.570.361.5216.1
Logit MagnitudeBase Model=Llama3, LLM Size=8B2026.0565.154.861.220.4
MetaUEBase Model=Llama3, LLM Size=8B2026.0564.354.160.20
LogTokUBase Model=Llama3, LLM Size=8B2026.0563.151.860.720.4
LogTokUBase Model=Llama3, LLM Size=3B2026.0561.651.86027.4
MetaUEBase Model=Llama3, LLM Size=3B2026.0560.953.157.70
Self-ConsistencyLLM=Qwen3.5-4B2026.0560.150.656.7430.2
LogTokULLM=Qwen3.5, Size=35B*2026.0558.34361.338
LogTokULLM=Gemma4-4B2026.0552.460.754.421.5
LogTokULLM=Gemma4, Size=31B2026.0551.251.457.219.5
Hidden-State ScoreLLM=Gemma4-4B2026.0537.351.550.621.5