Dialog State Tracking on MultiWOZ 2.1 (test)
79.2Joint Goal AccuracySUMBT (Coverage)
Evaluation Results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SUMBT (Coverage)Training=Zero-shot2021.05 | 79.2 | 60.4 | 88.6 | 76.2 | 95 | 74.5 | 88.7 | 86.9 | 97.3 | 97.8 | 97.6 | 93.4 | — | — | |
| TRADE (Coverage)Training=Zero-shot2021.05 | 66.9 | 52.5 | 82.2 | 68 | 90.8 | 59.1 | 75 | 65 | 89.3 | 90 | 89.9 | 85.4 | — | — | |
| SUMBTTraining=Full dataset2021.05 | 66.5 | 51.8 | 92.2 | 64.2 | 93.1 | 71.1 | 89.1 | 77 | 95 | 68.2 | 86 | 91.1 | — | — | |
| TRADETraining=Full dataset2021.05 | 65.3 | 50.5 | 91.4 | 61.8 | 92.7 | 67.3 | 87.6 | 74 | 94 | 72.7 | 88.9 | 89.8 | — | — | |
| TripPyType=Open vocabulary2022.03 | 61.31 | — | — | — | — | — | — | — | — | — | — | 97.07 | 85.73 | 84.32 | |
| DiCoS-DSTk=22022.05 | 61.02 | — | — | — | — | — | — | — | — | — | — | 98.05 | — | — | |
| DiCoS-DSTk=12022.05 | 60.89 | — | — | — | — | — | — | — | — | — | — | 98.05 | — | — | |
| DiCoS-DSTk=32022.05 | 60.85 | — | — | — | — | — | — | — | — | — | — | 98.05 | — | — | |
| DSS-DST2022.05 | 60.73 | — | — | — | — | — | — | — | — | — | — | 98.05 | — | — | |
| FPDSC-dualApproach Category=Classification-based, Requires Full Ontology=true2021.09 | 59.07 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CHANApproach Category=Classification-based, Requires Full Ontology=true2021.09 | 58.55 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FPDSC-turnApproach Category=Classification-based, Requires Full Ontology=true2021.09 | 57.88 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SimpleTODDecoder=Generative, Context Encoder=Unidirectional, Label cleaning=Type 2 and partial Type 42020.05 | 57.47 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PPTODlargeApproach Category=Generation-based, Requires Full Ontology=false2021.09 | 57.45 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PPTODbaseApproach Category=Generation-based, Requires Full Ontology=false2021.09 | 57.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PPTODpreprocessing=TripPy preprocessed2022.04 | 57.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SOLOISTApproach Category=Generation-based, Requires Full Ontology=false2021.09 | 56.85 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SOLOIST2020.05 | 56.85 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DSGFNetpreprocessing=TripPy preprocessed2022.04 | 56.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DST-as-Prompting2022.05 | 56.66 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PPTODsmallApproach Category=Generation-based, Requires Full Ontology=false2021.09 | 56.47 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| UBARApproach Category=Generation-based, Requires Full Ontology=false2021.09 | 56.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Seq2seq-DUApproach Category=Generation-based, Requires Full Ontology=false2021.09 | 56.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Seq2seq-DU2022.05 | 56.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Simple-TODType=Open vocabulary2022.03 | 56.05 | — | — | — | — | — | — | — | — | — | — | 97.61 | 92.76 | 87.97 | |
| SimpleTODApproach Category=Generation-based, Requires Full Ontology=false2021.09 | 55.76 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SimpleTODDecoder=Generative, Context Encoder=Unidirectional, Label cleaning=Wu et al. [53]2020.05 | 55.76 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SimpleTODDecoder=Generative, Context Encoder=Unidirectional, Label cleaning=none2020.05 | 55.72 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Simple-TOD2020.05 | 55.72 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TripPyDecoder=Classifier, Context Encoder=Bidirectional, Extra Supervision=action decision, Label cleaning=Heck et al. [19]2020.05 | 55.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TripPypreprocessing=TripPy preprocessed2022.04 | 55.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TripPyApproach Category=Classification-based, Requires Full Ontology=false2021.09 | 55.29 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TripPy2022.05 | 55.29 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Tripy2020.05 | 55.29 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SSTApproach Category=Classification-based, Requires Full Ontology=true2021.09 | 55.23 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SSTDecoder=Generative, Context Encoder=Bidirectional, Extra Supervision=schema graph, Label cleaning=Wu et al. [53]2020.05 | 55.23 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SST2020.05 | 55.23 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SSTpreprocessing=Original MultiWOZ2022.04 | 55.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DST-STARType=Ontology based2022.03 | 54.83 | — | — | — | — | — | — | — | — | — | — | 97.54 | 92.53 | 87.8 | |
| Transformer-DSTType=Open vocabulary2022.03 | 54.46 | — | — | — | — | — | — | — | — | — | — | 97.48 | 92.29 | 87.59 | |
| Pegasus-DST2022.05 | 54.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MinTLApproach Category=Generation-based, Requires Full Ontology=false2021.09 | 53.62 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MinTL2020.05 | 53.62 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MinTL-BARTpreprocessing=TripPy preprocessed2022.04 | 53.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SAVNType=Open vocabulary2022.03 | 53.57 | — | — | — | — | — | — | — | — | — | — | 97.49 | 92.46 | 87.69 | |
| SUMBT + NeuralWOZTraining=Zero-shot2021.05 | 53.4 | 31.3 | 81.7 | 48.9 | 88.4 | 53 | 79 | 66.9 | 92.4 | 66.7 | 83.9 | 85.1 | — | — | |
| DS-PicklistApproach Category=Classification-based, Requires Full Ontology=true2021.09 | 53.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DST-PicklistDecoder=Classifier, Context Encoder=Bidirectional, Label cleaning=Wu et al. [53]2020.05 | 53.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SOM-DSTpreprocessing=TripPy preprocessed2022.04 | 53 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SOM-DSTApproach Category=Generation-based, Requires Full Ontology=false2021.09 | 52.57 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SOM-DST2020.05 | 52.57 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SOM-DSTType=Open vocabulary2022.03 | 52.42 | — | — | — | — | — | — | — | — | — | — | 97.35 | 91.79 | 86.95 | |
| L4P4K2-DSGraphType=Ontology based2022.03 | 51.78 | — | — | — | — | — | — | — | — | — | — | 96.9 | 91.89 | 85.7 | |
| DS-DSTpreprocessing=TripPy preprocessed2022.04 | 51.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DSTQAApproach Category=Generation-based, Requires Full Ontology=true2021.09 | 51.17 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DSTQADecoder=Classifier, Context Encoder=Bidirectional, Extra Supervision=knowledge graph, Label cleaning=Wu et al. [53] + label variants2020.05 | 51.17 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DSTQA2020.05 | 51.17 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LABES-S2SApproach Category=Generation-based, Requires Full Ontology=false2021.09 | 51.15 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TRADEType=Open vocabulary2022.03 | 49.39 | — | — | — | — | — | — | — | — | — | — | 97 | 90.33 | 85.2 | |
| NADST2020.05 | 49.04 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| COMERType=Open vocabulary2022.03 | 48.79 | — | — | — | — | — | — | — | — | — | — | 96.52 | 88 | 82.5 | |
| SUMBT + ATDMTraining=Zero-shot2021.05 | 48.7 | 36.3 | 83.7 | 45.3 | 82.8 | 52.8 | 78.9 | 46.7 | 84.2 | 62.6 | 79.4 | 81.8 | — | — | |
| TOD-BERTApproach Category=Classification-based, Requires Full Ontology=false2021.09 | 48 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TOD-BERT2020.05 | 48 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SUMBTType=Ontology based2022.03 | 46.99 | — | — | — | — | — | — | — | — | — | — | 96.66 | 89.34 | 83.8 | |
| TRADEApproach Category=Generation-based, Requires Full Ontology=false2021.09 | 46 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TRADEpreprocessing=Original MultiWOZ2022.04 | 46 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TRADEDecoder=Generative + Classifier, Context Encoder=Bidirectional, Label cleaning=Wu et al. [53]2020.05 | 45.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TRADE2022.05 | 45.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TRADE2020.05 | 45.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TRADE + NeuralWOZTraining=Zero-shot2021.05 | 44.4 | 26.5 | 75.1 | 42 | 84.2 | 39.8 | 65.7 | 48.1 | 83.9 | 65.4 | 79.9 | 77.8 | — | — | |
| SGD-baselinepreprocessing=TripPy preprocessed2022.04 | 43.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TRADE + ATDMTraining=Zero-shot2021.05 | 40.3 | 28.3 | 74.5 | 35.9 | 75.6 | 34.9 | 62.2 | 37.4 | 74.5 | 65 | 79.9 | 73.3 | — | — | |
| DualLoRAPre. Model=PPTOD-small, Zero-shot=true2024.07 | 38.7 | 18.9 | — | 27.9 | — | 37.1 | — | 42.4 | — | 67.2 | — | — | — | — | |
| FJSTApproach Category=Classification-based, Requires Full Ontology=false2021.09 | 38 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FJST2020.05 | 38 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DCCPre. Model=T5-small, Zero-shot=true2024.07 | 37.9 | 24.8 | — | 22.9 | — | 35.8 | — | 40.2 | — | 65.9 | — | — | — | — | |
| PrompterPre. Model=PPTOD-small, Zero-shot=true2024.07 | 37.2 | 19.2 | — | 26 | — | 35.8 | — | 39 | — | 66.3 | — | — | — | — | |
| T5DST*Pre. Model=PPTOD-small, Zero-shot=true2024.07 | 36.4 | 20 | — | 25.3 | — | 35.5 | — | 35.3 | — | 65.6 | — | — | — | — | |
| SlotDM-DSTPre. Model=T5-small, Zero-shot=true2024.07 | 35.35 | 18.85 | — | 20.75 | — | 33.92 | — | 36.96 | — | 66.25 | — | — | — | — | |
| T5DSTPre. Model=T5-small, Zero-shot=true2024.07 | 33.14 | 20.72 | — | 20.09 | — | 31.92 | — | 28.83 | — | 64.12 | — | — | — | — | |
| GPT2-DSTPre. Model=GPT2-base, Zero-shot=true2024.07 | 29.34 | 18.54 | — | 21.05 | — | 23.67 | — | 24.34 | — | 59.1 | — | — | — | — | |
| TRADE (Campagna)Training=Zero-shot2021.05 | 28.2 | 19.5 | 62.6 | 16.4 | 51.5 | 22.8 | 50 | 22.9 | 48 | 59.2 | 72 | 56.8 | — | — | |
| SUMBTTraining=Zero-shot2021.05 | 28.2 | 19.8 | 63.3 | 16.5 | 52.1 | 22.6 | 51.5 | 22.5 | 49.2 | 59.5 | 74.9 | 58.2 | — | — | |
| SUMBTPre. Model=BERT-base, Zero-shot=true2024.07 | 28.04 | 19.08 | — | 16.5 | — | 22.6 | — | 22.5 | — | 59.5 | — | — | — | — | |
| MA-DSTZero-shot=true2024.07 | 26.87 | 16.28 | — | 13.56 | — | 22.46 | — | 22.76 | — | 59.27 | — | — | — | — | |
| TRADE (Wu)Training=Zero-shot2021.05 | 25.8 | 13.7 | 65.6 | 13.4 | 54.5 | 20.5 | 55.5 | 21 | 48.9 | 60.2 | 73.5 | 59.6 | — | — | |
| TRADEZero-shot=true2024.07 | 25.69 | 14.2 | — | 12.59 | — | 20.06 | — | 22.39 | — | 59.21 | — | — | — | — |