List your company

Confidence Estimation on Open-Set

67.2AccuracyGPT-4.1 + Verbalized Conf.
28.61638.63348.6558.667Nov 18, 2025

Evaluation results

MethodLinks
GPT-4.1 + Verbalized Conf.Model=GPT-4.1, Verbalization Strategy=Verbalized Confidence2025.1167.267.928.827.7
GPT-4.1 + Verbalized Probability DistributionModel=GPT-4.1, Verbalization Strategy=Verbalized Probability Distribution2025.1166.971.82110
GPT-4.1Model=GPT-4.1, Verbalization Strategy=None (Baseline)2025.1166.5———
GPT-4.1 + Verbalized Top-kModel=GPT-4.1, Verbalization Strategy=Verbalized Top-k2025.1164.870.52418.6
DeepSeek-V3 + Verbalized Conf.Model=DeepSeek-V3, Verbalization Strategy=Verbalized Confidence2025.1158.371.230.128.5
DeepSeek-V3Model=DeepSeek-V3, Verbalization Strategy=None (Baseline)2025.1157.6———
DeepSeek-V3 + Verbalized Probability DistributionModel=DeepSeek-V3, Verbalization Strategy=Verbalized Probability Distribution2025.1157.373.121.59.4
DeepSeek-V3 + Verbalized Top-kModel=DeepSeek-V3, Verbalization Strategy=Verbalized Top-k2025.1155.867.128.424.2
Qwen3-30B-A3B-Instruct + Verbalized Top-kModel=Qwen3-30B-A3B-Instruct, Verbalization Strategy=Verbalized Top-k2025.1144.463.446.847.7
Qwen3-30B-A3B-Instruct + Verbalized Conf.Model=Qwen3-30B-A3B-Instruct, Verbalization Strategy=Verbalized Confidence2025.1143.562.651.251.9
Qwen3-30B-A3B-Instruct + Verbalized Probability DistributionModel=Qwen3-30B-A3B-Instruct, Verbalization Strategy=Verbalized Probability Distribution2025.1143.468.130.427.4
Qwen3-30B-A3B-InstructModel=Qwen3-30B-A3B-Instruct, Verbalization Strategy=None (Baseline)2025.1142.1———
Qwen3-4B-Instruct + Verbalized Probability DistributionModel=Qwen3-4B-Instruct, Verbalization Strategy=Verbalized Probability Distribution2025.1132.468.133.333.1
Qwen3-4B-Instruct + Verbalized Top-kModel=Qwen3-4B-Instruct, Verbalization Strategy=Verbalized Top-k2025.11316453.957
Qwen3-4B-Instruct + Verbalized Conf.Model=Qwen3-4B-Instruct, Verbalization Strategy=Verbalized Confidence2025.1130.364.560.963
Qwen3-4B-InstructModel=Qwen3-4B-Instruct, Verbalization Strategy=None (Baseline)2025.1130.1———