List your company
ResearchFormerly WizWand

Correctness Prediction on Math operations

0.913AUROCVerb. conf.
0.313960.469480.6250.78052Sep 12, 2025
Updated 8mo ago

Evaluation results

MethodLinks
Verb. conf.Model=Llama 3.3 70B Instruct2025.090.913
DirectionModel=Llama 3.1 8B, Train dataset=TriviaQA2025.090.858
DirectionModel=DeepSeek R1 Distill Qwen 32B, Train dataset=TriviaQA2025.090.847
DirectionModel=Ministral 8B Instruct 2410, Train dataset=TriviaQA2025.090.844
DirectionModel=Qwen 2.5 7B Instruct, Train dataset=TriviaQA2025.090.837
DirectionModel=Llama 3.3 70B Instruct, Train dataset=TriviaQA2025.090.835
DirectionModel=Mistral 7B Instruct v0.3, Train dataset=TriviaQA2025.090.782
Verb. conf.Model=Llama 3.1 8B2025.090.623
Verb. conf.Model=Mistral 7B Instruct v0.32025.090.617
AssessorModel=Llama 3.1 8B, Assessor type=logistic regression2025.090.528
Verb. conf.Model=Qwen 2.5 7B Instruct2025.090.517
Verb. conf.Model=Ministral 8B Instruct 24102025.090.5
Verb. conf.Model=DeepSeek R1 Distill Qwen 32B2025.090.499
AssessorModel=Mistral 7B Instruct v0.3, Assessor type=logistic regression2025.090.493
AssessorModel=Ministral 8B Instruct 2410, Assessor type=logistic regression2025.090.454
AssessorModel=Llama 3.3 70B Instruct, Assessor type=logistic regression2025.090.449
AssessorModel=Qwen 2.5 7B Instruct, Assessor type=logistic regression2025.090.4
AssessorModel=DeepSeek R1 Distill Qwen 32B, Assessor type=logistic regression2025.090.337