List your company

Confidence Calibration on MACE (test)

81.2AUROCSCA
47.08855.94464.873.656Feb 8, 2026

Evaluation results

MethodLinks
SCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=0.052026.0281.2
SCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=02026.0281.2
PerplexityBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0280.9
SCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=0.052026.0280.6
SCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=02026.0280.5
ConsisBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0280.3
ConsisBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0280.2
Consis-Verb-TopkBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0280
Consis-VerbBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0279.7
PerplexityBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0279.4
SCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=0.052026.0279.2
Consis-VerbBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0279.1
SCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=02026.0279.1
Consis-Verb-TopkBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0279
Sem EntropyBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Question-level2026.0278.9
Sem EntropyBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Question-level2026.0278.6
ConsisBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0278.1
SNCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=0.52026.0277.9
SNCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=0.52026.0277.3
SFCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=0.252026.0277.2
SFCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=0.252026.0277.1
Consis-Verb-TopkBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0277
SCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=0.052026.0276.7
SCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=02026.0276.7
Consis-VerbBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0276.6
PerplexityBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0276.5
PTrue-Prob-CandBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0276
Sem EntropyBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Question-level2026.0275.8
ConsisBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0275.3
SFCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=0.252026.0275.2
Prob EntropyBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Question-level2026.0274.5
Consis-Verb-TopkBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0274.3
SNCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=0.52026.0274.2
N-Prob EntropyBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Question-level2026.0273.8
Consis-VerbBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0273.8
PerplexityBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0273.8
PTrue-Prob-CandBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0273.5
SFCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=0.252026.0273
PTrue-ProbBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0272.9
Prob EntropyBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Question-level2026.0272.6
PTrue-Prob-CandBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0272.5
Sem EntropyBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Question-level2026.0272.5
PTrue-Prob-CandBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0271.1
SNCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=0.52026.0270.5
N-Prob EntropyBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Question-level2026.0270.4
PTrue-ProbBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0270.1
Prob EntropyBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Question-level2026.0269
PTrue-ConsisBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0268.1
PTrue-ProbBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0267.6
Self-Ask-CandBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0266.9
Self-Ask-CandBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0266.8
PTrue-ProbBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0266.7
Self-AskBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0266.5
Self-Ask-CandBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0266.5
PTrue-ConsisBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0266.4
N-Prob EntropyBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Question-level2026.0266.2
Self-AskBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0266.1
Prob EntropyBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Question-level2026.0266.1
Self-Ask-CandBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0266.1
Self-AskBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0265
PTrue-ConsisBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0264.1
Self-AskBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0263.9
PTrue-ConsisBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0263.5
N-Prob EntropyBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Question-level2026.0263.3
PTrue-Consis-CandBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0262.9
PTrue-Consis-CandBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0262.7
PTrue-Consis-CandBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0262.6
PTrue-Consis-CandBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Double-turn2026.0262.5
Verb-TopkBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0260.7
VerbBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0258.4
VerbBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0258.4
VerbBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0257.2
Verb-TopkBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0256.2
VerbBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0255.8
Verb-TopkBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0251.2
SNCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=02026.0250
SFCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=02026.0250
SNCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=02026.0250
SFCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=02026.0250
SNCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=02026.0250
SFCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=02026.0250
SNCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=02026.0250
SFCABackbone=LLaMA-3.1-70B-Instruct, Filtering Threshold (τ)=02026.0250
Verb-TopkBackbone=LLaMA-3.1-70B-Instruct, Interaction Turn=Single-turn, Calibration Granularity=Answer-level2026.0248.4