List your company
ResearchFormerly WizWand

Question Answering on MedQA (Accuracy)

71.72AccuracyAll
14.707229.508644.3159.1114Apr 7, 2026Apr 10, 2026Apr 14, 2026Apr 17, 2026Apr 21, 2026Apr 24, 2026Apr 28, 2026
Updated 6mo ago

Evaluation results

MethodLinks
AllIteration=12026.0471.72
EVOSELECTIteration=2, Selection Ratio=0.22026.0471.64
AttributionIteration=1, Selection Ratio=0.52026.0471.41
TSDSIteration=1, Selection Ratio=0.22026.0471.35
RandomIteration=2, Selection Ratio=0.22026.0471.33
EVOSELECTIteration=1, Selection Ratio=0.52026.0471.09
EVOSELECTIteration=1, Selection Ratio=0.22026.0471.01
Attr-DivIteration=2, Selection Ratio=0.22026.0471.01
AllIteration=22026.0470.93
AttributionIteration=1, Selection Ratio=0.22026.0470.86
AttributionIteration=2, Selection Ratio=0.22026.0470.86
EVOSELECTIteration=2, Selection Ratio=0.52026.0470.86
Attr-DivIteration=2, Selection Ratio=0.52026.0470.78
RandomIteration=1, Selection Ratio=0.22026.0470.62
RandomIteration=1, Selection Ratio=0.52026.0470.62
RandomIteration=2, Selection Ratio=0.52026.0470.62
DiversityIteration=1, Selection Ratio=0.22026.0470.54
Attr-DivIteration=1, Selection Ratio=0.22026.0470.54
DiversityIteration=2, Selection Ratio=0.22026.0470.54
Attr-DivIteration=1, Selection Ratio=0.52026.0470.38
TSDSIteration=2, Selection Ratio=0.22026.0470.38
AttributionIteration=2, Selection Ratio=0.52026.0470.31
DiversityIteration=2, Selection Ratio=0.52026.0470.31
Base2026.0470.23
DiversityIteration=1, Selection Ratio=0.52026.0469.99
TSDSIteration=1, Selection Ratio=0.52026.0469.91
TSDSIteration=2, Selection Ratio=0.52026.0469.68
POPBackbone=Qwen-2.5-7B-Inst, Evaluation Protocol=0-shot2026.0456.47
POPModel=Qwen-2.5-7B-Inst, Training Strategy=POP, Evaluation Protocol=0-shot2026.0456.08
BaseBackbone=Qwen-2.5-7B-Inst, Evaluation Protocol=0-shot2026.0455.94
BaseModel=Qwen-2.5-7B-Inst, Training Strategy=Base, Evaluation Protocol=0-shot2026.0455.94
Train on DBackbone=Qwen-2.5-7B-Inst, Evaluation Protocol=0-shot2026.0455.3
Train on DBackbone=Qwen-2.5-7B, Evaluation Protocol=0-shot2026.0453.65
EVOSELECTIteration=1, Selection Ratio=0.5, Backbone=3b base model2026.0453.34
Train on DModel=Qwen-2.5-7B-Inst, Training Strategy=Naive pretraining on D, Evaluation Protocol=0-shot2026.0453.33
Attr-DivIteration=1, Selection Ratio=0.2, Backbone=3b base model2026.0453.02
POPBackbone=Qwen-2.5-7B, Evaluation Protocol=0-shot2026.0452.98
BaseBackbone=Qwen-2.5-7B, Evaluation Protocol=0-shot2026.0452.95
BaseModel=Qwen-2.5-7B, Training Strategy=Base, Evaluation Protocol=0-shot2026.0452.95
POPModel=Qwen-2.5-7B, Training Strategy=POP, Evaluation Protocol=0-shot2026.0452.95
RandomIteration=1, Selection Ratio=0.2, Backbone=3b base model2026.0452.63
AttributionIteration=1, Selection Ratio=0.2, Backbone=3b base model2026.0452.32
DiversityIteration=2, Selection Ratio=0.2, Backbone=3b base model2026.0452.24
RandomIteration=1, Selection Ratio=0.5, Backbone=3b base model2026.0452.08
DiversityIteration=1, Selection Ratio=0.5, Backbone=3b base model2026.0451.85
EVOSELECTIteration=1, Selection Ratio=0.2, Backbone=3b base model2026.0451.69
EVOSELECTIteration=2, Selection Ratio=0.5, Backbone=3b base model2026.0451.61
EVOSELECTIteration=2, Selection Ratio=0.2, Backbone=3b base model2026.0451.53
TSDSIteration=2, Selection Ratio=0.2, Backbone=3b base model2026.0451.45
TSDSIteration=1, Selection Ratio=0.2, Backbone=3b base model2026.0451.37
Attr-DivIteration=1, Selection Ratio=0.5, Backbone=3b base model2026.0451.06
AttributionIteration=2, Selection Ratio=0.5, Backbone=3b base model2026.0450.82
AttributionIteration=1, Selection Ratio=0.5, Backbone=3b base model2026.0450.67
TSDSIteration=2, Selection Ratio=0.5, Backbone=3b base model2026.0450.67
BaseIteration=Base, Backbone=3b base model2026.0450.43
TSDSIteration=1, Selection Ratio=0.5, Backbone=3b base model2026.0450.43
DiversityIteration=1, Selection Ratio=0.2, Backbone=3b base model2026.0450.35
AllIteration=2, Selection Ratio=All, Backbone=3b base model2026.0450.35
RandomIteration=2, Selection Ratio=0.2, Backbone=3b base model2026.0450.35
Attr-DivIteration=2, Selection Ratio=0.2, Backbone=3b base model2026.0450.27
RandomIteration=2, Selection Ratio=0.5, Backbone=3b base model2026.0450.2
AttributionIteration=2, Selection Ratio=0.2, Backbone=3b base model2026.0450.2
DiversityIteration=2, Selection Ratio=0.5, Backbone=3b base model2026.0449.65
Attr-DivIteration=2, Selection Ratio=0.5, Backbone=3b base model2026.0449.33
AllIteration=1, Selection Ratio=All, Backbone=3b base model2026.0449.18
Train on DModel=Qwen-2.5-7B, Training Strategy=Naive pretraining on D, Evaluation Protocol=0-shot2026.0449.13
T5 (beam)QA Model=T52026.0432.68
T5 (beam)QA Model=BART2026.0431.81
GPT-3 (clustering)QA Model=T52026.0429.69
GPT-3 (clustering)QA Model=BART2026.0428.04
T5 (contrast)QA Model=T52026.0426.71
TinyLlamaQA Model=BART2026.0425.22
T5 (contrast)QA Model=BART2026.0424.19
TinyLlamaQA Model=T52026.0424.12
Ground-TruthQA Model=T52026.0421.76
GPT-3 (k-NN)QA Model=T52026.0421.05
Ground-TruthQA Model=BART2026.0420.42
SFT2026.0420.4
GPT-3 (k-NN)QA Model=BART2026.0420.03
GPT-3 (COT-k-NN)QA Model=T52026.0419.95
R-Tuning2026.0419.7
GPT-3 (COT-k-NN)QA Model=BART2026.0419.25
KWT (LLM)Knowledge Estimation=LLM2026.0418.5
FT-TOP2026.0417.2
KWT (EM)Knowledge Estimation=EM2026.0416.9
KWT (Rouge)Knowledge Estimation=Rouge2026.0416.9