List your company

Knowledge-grounded Dialog Generation on Wizard of Wikipedia (WoW)

97.7Win RatePOKI
20.84440.79760.7580.703Mar 22, 2022

Evaluation results

MethodLinks
POKIComparison Baseline=POKI w/o Entailment, Criteria=Engaging2022.0397.70.80.7
POKIComparison Baseline=RAG, Criteria=Interesting2022.0396.33.50.74
POKIComparison Baseline=RAG, Criteria=Coherent2022.0395.44.50.78
POKIComparison Baseline=POKI w/o Entailment, Criteria=Coherent2022.0394.33.40.68
POKIComparison Baseline=POKI w/o Kw Fidelity, Criteria=Interesting2022.0393.54.50.71
POKIComparison Baseline=BART, Criteria=Humanlike2022.0392.46.50.66
POKIComparison Baseline=RAG, Criteria=Humanlike2022.0391.47.10.68
POKIComparison Baseline=RAG, Criteria=Engaging2022.0389.37.70.72
POKIComparison Baseline=BART, Criteria=Coherent2022.0388.59.60.72
POKIComparison Baseline=BART, Criteria=Engaging2022.0387.88.30.71
POKIComparison Baseline=POKI w/o Entailment, Criteria=Humanlike2022.0384.510.50.67
POKIComparison Baseline=POKI w/o Kw Fidelity, Criteria=Coherent2022.0383.610.70.65
POKIComparison Baseline=BART, Criteria=Interesting2022.0383.39.90.75
POKIComparison Baseline=POKI w/o Kw Fidelity, Criteria=Humanlike2022.0381.813.50.71
POKIComparison Baseline=POKI w/o Entailment, Criteria=Interesting2022.0379.817.20.7
POKIComparison Baseline=POKI w/o Kw Fidelity, Criteria=Engaging2022.0371.525.40.69
POKIComparison Baseline=Gold, Criteria=Humanlike2022.0342.341.90.68
POKIComparison Baseline=Gold, Criteria=Interesting2022.0335.937.80.76
POKIComparison Baseline=Gold, Criteria=Engaging2022.0325.426.70.73
POKIComparison Baseline=Gold, Criteria=Coherent2022.0323.825.30.73