List your company

Calibration on SQuAD

5.87ECETemp. Scaling
3.0921.85540.6259.385Jan 6, 2026Jan 12, 2026Jan 18, 2026Jan 25, 2026Jan 31, 2026Feb 6, 2026Feb 13, 2026

Evaluation results

MethodLinks
Temp. ScalingModel=Olmo2-7B-Instruct, Unsupervised=false2026.015.870.1769
Temp. ScalingModel=Llama3.1-8B-Instruct, Unsupervised=false2026.019.110.1818
Temp. ScalingModel=Qwen2.5-7B-Instruct, Unsupervised=false2026.019.660.2007
BaseCal-ReEvalModel=Olmo2-7B-Instruct, Unsupervised=true2026.0121.310.2216
BaseCal-ProjModel=Olmo2-7B-Instruct, Unsupervised=true2026.0123.930.2357
VerbalizationModel=Olmo2-7B-Instruct, Unsupervised=true2026.0129.560.304
BaseCal-ReEvalModel=Llama3.1-8B-Instruct, Unsupervised=true2026.0129.590.2755
CDKCBackbone=Qwen2.5-7B-Instruct, Optimization Rounds=22026.0230.29—
BaseCal-ProjModel=Llama3.1-8B-Instruct, Unsupervised=true2026.0131.340.2816
BaseCal-ReEvalModel=Qwen2.5-7B-Instruct, Unsupervised=true2026.0132.150.298
CDKCBackbone=Qwen2.5-7B-Instruct2026.0235.69—
VerbalizationModel=Llama3.1-8B-Instruct, Unsupervised=true2026.0136.030.3208
GRPOBackbone=Qwen2.5-7B-Instruct2026.0237.92—
Semantic EntropyModel=Olmo2-7B-Instruct, Unsupervised=true2026.0139.060.3558
Semantic EntropyModel=Qwen2.5-7B-Instruct, Unsupervised=true2026.0140.410.3821
BaseCal-ProjModel=Qwen2.5-7B-Instruct, Unsupervised=true2026.0140.850.3569
VanillaModel=Olmo2-7B-Instruct, Unsupervised=true2026.0144.590.3733
Semantic EntropyModel=Llama3.1-8B-Instruct, Unsupervised=true2026.0146.450.4003
VerbalizationModel=Qwen2.5-7B-Instruct, Unsupervised=true2026.01480.4646
P(true)Model=Olmo2-7B-Instruct, Unsupervised=true2026.0148.220.4582
VanillaModel=Llama3.1-8B-Instruct, Unsupervised=true2026.0152.550.4546
P(true)Model=Llama3.1-8B-Instruct, Unsupervised=true2026.0155.320.5094
P(true)Model=Qwen2.5-7B-Instruct, Unsupervised=true2026.0156.660.565
VanillaModel=Qwen2.5-7B-Instruct, Unsupervised=true2026.0160.120.555
BARRELBackbone=Qwen2.5-7B-Instruct2026.0265.13—
Know WhatBackbone=Qwen2.5-7B-Instruct2026.0267.03—
Vanilla LLMBackbone=Qwen2.5-7B-Instruct2026.0269.23—
Vanilla SFTBackbone=Qwen2.5-7B-Instruct2026.0269.84—
LLKD-SFTBackbone=Qwen2.5-7B-Instruct2026.0271.13—
CGKEBackbone=Qwen2.5-7B-Instruct2026.0271.24—
CRew-DPOBackbone=Qwen2.5-7B-Instruct2026.0275.37—