List your company

Uncertainty Estimation on NewsQA

76.6AUROCSelf-Consistency
44.98453.19261.469.608May 7, 2026

Evaluation results

MethodLinks
Self-ConsistencyLLM=Qwen3.5-4B2026.0576.680.770.2161.1
Semantic EntropyLLM=Qwen3.5-4B2026.0576.179.871.4161.1
Logit MagnitudeLLM=Gemma4, Size=31B2026.0575.984.669.78.9
LogTokUBase Model=Llama3, LLM Size=8B2026.0575.763.368.327.2
Logit MagnitudeLLM=Gemma4, Size=26B*2026.0575.583.369.29.7
Logit MagnitudeLLM=Qwen3.5, Size=9B2026.0575.377.268.624.1
Logit MagnitudeLLM=Qwen3.5-4B, generation=Full (τ = T)2026.0575.179.470.515.7
Logit MagnitudeLLM=Qwen3.5-4B, generation=Early-stopped2026.057579.469.911.7
LogTokULLM=Qwen3.5, Size=9B2026.0574.974.468.924.1
LogTokULLM=Qwen3.5-4B2026.0574.277.969.615.7
Logit MagnitudeBase Model=Llama3, LLM Size=8B2026.0573.963.468.723.1
LogTokUBase Model=Llama3, LLM Size=3B2026.0572.853.266.629
Logit MagnitudeBase Model=Llama3, LLM Size=3B2026.0572.75367.329
LogTokULLM=Qwen3.5, Size=35B*2026.0572.775.166.716.3
Logit MagnitudeLLM=Qwen3.5, Size=35B*2026.0572.277.467.116.3
LogTokULLM=Gemma4, Size=26B*2026.0571.981.765.79.7
MetaUELLM=Qwen3.5-4B2026.0571.778.865.60
MetaUELLM=Gemma4, Size=31B2026.0571.783.665.90
MetaUELLM=Gemma4, Size=26B*2026.0571.482.264.80
LogTokULLM=Gemma4-4B2026.0571.1656420.8
Logit MagnitudeLLM=Gemma4-4B, generation=Full (τ = T)2026.0570.465.665.520.8
Logit MagnitudeLLM=Gemma4-4B, generation=Early-stopped2026.0569.965.264.816
Hidden-State ScoreLLM=Gemma4-4B2026.0569.665.664.320.8
LogTokULLM=Gemma4, Size=31B2026.0569.482.364.78.9
MetaUELLM=Qwen3.5, Size=35B*2026.0568.676.363.70
MetaUELLM=Qwen3.5, Size=9B2026.0567.974.2630
Self-CertaintyLLM=Gemma4, Size=31B2026.0566.381.360.88.9
Hidden-State ScoreLLM=Qwen3.5-4B2026.0565.174.559.815.7
Sequence EntropyLLM=Qwen3.5-4B2026.0564.875.660.515.7
Self-CertaintyBase Model=Llama3, LLM Size=8B2026.0563.456.25927.2
Self-CertaintyLLM=Gemma4, Size=26B*2026.0563.278.259.29.7
MetaUELLM=Gemma4-4B2026.0561.759.957.90
Semantic EntropyLLM=Gemma4-4B2026.0561.258.358.5209.7
Self-ConsistencyLLM=Gemma4-4B2026.0561.258.557.7209.7
Self-CertaintyBase Model=Llama3, LLM Size=3B2026.0560.345.257.229
MetaUEBase Model=Llama3, LLM Size=8B2026.0558.252.955.70
MetaUEBase Model=Llama3, LLM Size=3B2026.0557.743.855.10
Self-CertaintyLLM=Qwen3.5-4B2026.0556.671.354.815.7
Self-CertaintyLLM=Qwen3.5, Size=35B*2026.0555.369.253.416.3
Self-CertaintyLLM=Qwen3.5, Size=9B2026.0554.866.953.424.1
Self-CertaintyLLM=Gemma4-4B2026.0551.154.253.520.8
Sequence EntropyLLM=Gemma4-4B2026.0546.250.551.120.8