Response Selection on Douban Conversation Corpus (test)
0.675MAPBERT_TL
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| BERT_TLModel Type=Our Model2022.03 | 0.675 | 0.718 | 0.564 | — | — | — | 0.367 | 0.571 | 0.874 | 0.981 | |
| Uni-Enc+BERT-FPProtocol=post-trained2021.06 | 0.648 | 0.688 | 0.518 | — | — | — | 0.327 | 0.557 | 0.865 | — | |
| BERT-FPModel Type=PrLMs-based2022.03 | 0.644 | 0.68 | 0.512 | — | — | — | 0.324 | 0.542 | 0.87 | 0.911 | |
| BERT-FPProtocol=post-trained2021.06 | 0.644 | 0.68 | 0.512 | — | — | — | 0.324 | 0.542 | 0.87 | — | |
| SA-BERT+HCLTraining=Hierarchical Curriculum Learning2020.12 | 0.639 | 0.681 | 0.514 | — | — | — | 0.33 | 0.531 | 0.858 | — | |
| SA-BERT+HCLProtocol=fine-tuned2021.06 | 0.639 | 0.681 | 0.514 | — | — | — | 0.33 | 0.531 | 0.858 | — | |
| BERT-UMS+FGCProtocol=post-trained2021.06 | 0.627 | 0.67 | 0.5 | — | — | — | 0.326 | 0.512 | 0.869 | — | |
| UMSBERT+Domain-specific post-training=true2020.09 | 0.625 | 0.664 | 0.499 | — | — | — | 0.318 | 0.482 | 0.858 | — | |
| UMS-BERT+Model Type=PrLMs-based2022.03 | 0.625 | 0.664 | 0.499 | — | — | — | 0.318 | 0.482 | 0.858 | 0.875 | |
| UMSBERT+Protocol=post-trained2021.06 | 0.625 | 0.664 | 0.499 | — | — | — | 0.318 | 0.482 | 0.858 | — | |
| MDFNModel Type=PrLMs-based2022.03 | 0.624 | 0.663 | 0.498 | — | — | — | 0.325 | 0.511 | 0.855 | 0.967 | |
| MDFNProtocol=fine-tuned2021.06 | 0.624 | 0.663 | 0.498 | — | — | — | 0.325 | 0.511 | 0.855 | — | |
| UMSELECTRA+Domain-specific post-training=true2020.09 | 0.623 | 0.663 | 0.492 | — | — | — | 0.307 | 0.501 | 0.851 | — | |
| UMS-ELECTRA+Model Type=PrLMs-based2022.03 | 0.623 | 0.663 | 0.492 | — | — | — | 0.307 | 0.501 | 0.851 | 0.875 | |
| Uni-EncoderProtocol=fine-tuned2021.06 | 0.622 | 0.662 | 0.481 | — | — | — | 0.303 | 0.514 | 0.852 | — | |
| MSN+HCLTraining=Hierarchical Curriculum Learning2020.12 | 0.62 | 0.668 | 0.507 | — | — | — | 0.321 | 0.508 | 0.841 | — | |
| SA-BERT2020.04 | 0.619 | 0.659 | 0.496 | 0.313 | 0.481 | 0.847 | — | — | — | — | |
| SA-BERTDomain-specific post-training=false2020.09 | 0.619 | 0.659 | 0.496 | — | — | — | 0.313 | 0.481 | 0.847 | — | |
| SA-BERT2020.12 | 0.619 | 0.659 | 0.496 | — | — | — | 0.313 | 0.481 | 0.847 | — | |
| SA-BERTModel Type=PrLM-based2022.03 | 0.619 | 0.659 | 0.496 | — | — | — | 0.313 | 0.481 | 0.847 | 0.965 | |
| SA-BERTProtocol=fine-tuned2021.06 | 0.619 | 0.659 | 0.496 | — | — | — | 0.313 | 0.481 | 0.847 | — | |
| BERT+FGCProtocol=fine-tuned2021.06 | 0.614 | 0.653 | 0.495 | — | — | — | 0.312 | 0.495 | 0.85 | — | |
| ELECTRA+Domain-specific post-training=true2020.09 | 0.612 | 0.655 | 0.48 | — | — | — | 0.301 | 0.499 | 0.836 | — | |
| BERT+Domain-specific post-training=true2020.09 | 0.609 | 0.645 | 0.463 | — | — | — | 0.29 | 0.505 | 0.838 | — | |
| UMSELECTRADomain-specific post-training=false2020.09 | 0.608 | 0.65 | 0.472 | — | — | — | 0.291 | 0.488 | 0.845 | — | |
| BERT-SS-DADomain-specific post-training=false2020.09 | 0.602 | 0.643 | 0.458 | — | — | — | 0.28 | 0.491 | 0.843 | — | |
| ELECTRADomain-specific post-training=false2020.09 | 0.602 | 0.642 | 0.465 | — | — | — | 0.287 | 0.483 | 0.839 | — | |
| G-MSNtraining protocol=grayscale enhanced training2020.04 | 0.599 | 0.645 | 0.476 | — | — | — | 0.308 | 0.468 | 0.826 | — | |
| ELECTRAModel Type=PrLMs-based2022.03 | 0.599 | 0.643 | 0.471 | — | — | — | 0.287 | 0.474 | 0.831 | 0.96 | |
| BERT (ours)Domain-specific post-training=false2020.09 | 0.597 | 0.634 | 0.448 | — | — | — | 0.279 | 0.489 | 0.823 | — | |
| UMSBERTDomain-specific post-training=false2020.09 | 0.597 | 0.639 | 0.466 | — | — | — | 0.285 | 0.471 | 0.829 | — | |
| UMSBERTProtocol=fine-tuned2021.06 | 0.597 | 0.639 | 0.466 | — | — | — | 0.285 | 0.471 | 0.829 | — | |
| BERT2020.04 | 0.591 | 0.633 | 0.454 | 0.28 | 0.47 | 0.828 | — | — | — | — | |
| G-IOItraining protocol=grayscale enhanced training2020.04 | 0.591 | 0.639 | 0.454 | — | — | — | 0.277 | 0.458 | 0.796 | — | |
| BERT (Gu et al. 2020)Domain-specific post-training=false2020.09 | 0.591 | 0.633 | 0.454 | — | — | — | 0.28 | 0.47 | 0.828 | — | |
| BERTModel Type=PrLM-based2022.03 | 0.591 | 0.633 | 0.454 | — | — | — | 0.28 | 0.47 | 0.828 | 0.95 | |
| BERTProtocol=fine-tuned2021.06 | 0.591 | 0.633 | 0.454 | — | — | — | 0.28 | 0.47 | 0.828 | — | |
| G-DAMtraining protocol=grayscale enhanced training2020.04 | 0.588 | 0.637 | 0.464 | — | — | — | 0.284 | 0.466 | 0.822 | — | |
| MSN2020.04 | 0.587 | 0.632 | 0.47 | 0.295 | 0.452 | 0.788 | — | — | — | — | |
| MSN2020.04 | 0.587 | 0.632 | 0.47 | — | — | — | 0.295 | 0.452 | 0.788 | — | |
| MSNDomain-specific post-training=false2020.09 | 0.587 | 0.632 | 0.47 | — | — | — | 0.295 | 0.452 | 0.788 | — | |
| MSN2020.12 | 0.587 | 0.632 | 0.47 | — | — | — | 0.295 | 0.452 | 0.788 | — | |
| MSNModel Type=non PrLM-based2022.03 | 0.587 | 0.632 | 0.47 | — | — | — | 0.295 | 0.452 | 0.788 | 0.8 | |
| SMN+HCLTraining=Hierarchical Curriculum Learning2020.12 | 0.575 | 0.62 | 0.446 | — | — | — | 0.281 | 0.452 | 0.807 | — | |
| IoI2020.04 | 0.573 | 0.621 | 0.444 | 0.269 | 0.451 | 0.786 | — | — | — | — | |
| IOI2020.04 | 0.573 | 0.621 | 0.444 | — | — | — | 0.269 | 0.451 | 0.786 | — | |
| IoIDomain-specific post-training=false2020.09 | 0.573 | 0.621 | 0.444 | — | — | — | 0.269 | 0.451 | 0.786 | — | |
| IOI2020.12 | 0.573 | 0.621 | 0.444 | — | — | — | 0.269 | 0.451 | 0.786 | — | |
| IoIModel Type=non PrLM-based2022.03 | 0.573 | 0.621 | 0.444 | — | — | — | 0.269 | 0.451 | 0.786 | 0.947 | |
| MRFN2020.04 | 0.571 | 0.617 | 0.448 | 0.276 | 0.435 | 0.783 | — | — | — | — | |
| MRFN2020.12 | 0.571 | 0.617 | 0.448 | — | — | — | 0.276 | 0.435 | 0.783 | — | |
| MRFNModel Type=non PrLM-based2022.03 | 0.571 | 0.617 | 0.448 | — | — | — | 0.276 | 0.435 | 0.783 | 0.945 | |
| IMN2020.04 | 0.57 | 0.615 | 0.433 | 0.262 | 0.452 | 0.789 | — | — | — | — | |
| IMNModel Type=non PrLM-based2022.03 | 0.57 | 0.615 | 0.433 | — | — | — | 0.262 | 0.452 | 0.789 | 0.794 | |
| G-SMNtraining protocol=grayscale enhanced training2020.04 | 0.564 | 0.615 | 0.443 | — | — | — | 0.271 | 0.439 | 0.781 | — | |
| DUA2020.04 | 0.551 | 0.599 | 0.421 | 0.243 | 0.421 | 0.78 | — | — | — | — | |
| DUA2020.04 | 0.551 | 0.599 | 0.421 | — | — | — | 0.243 | 0.421 | 0.78 | — | |
| DUADomain-specific post-training=false2020.09 | 0.551 | 0.599 | 0.421 | — | — | — | 0.243 | 0.421 | 0.78 | — | |
| DUA2020.12 | 0.551 | 0.599 | 0.421 | — | — | — | 0.243 | 0.421 | 0.78 | — | |
| DUAModel Type=non PrLM-based2022.03 | 0.551 | 0.599 | 0.421 | — | — | — | 0.243 | 0.421 | 0.78 | 0.752 | |
| DAM2020.04 | 0.55 | 0.601 | 0.427 | 0.254 | 0.41 | 0.757 | — | — | — | — | |
| DAM2020.04 | 0.55 | 0.601 | 0.427 | — | — | — | 0.254 | 0.41 | 0.757 | — | |
| DAMDomain-specific post-training=false2020.09 | 0.55 | 0.601 | 0.427 | — | — | — | 0.254 | 0.41 | 0.757 | — | |
| DAM2020.12 | 0.55 | 0.601 | 0.427 | — | — | — | 0.254 | 0.41 | 0.757 | — | |
| DAMModel Type=non PrLM-based2022.03 | 0.55 | 0.601 | 0.427 | — | — | — | 0.254 | 0.41 | 0.757 | 0.938 | |
| SMN2020.04 | 0.529 | 0.569 | 0.397 | 0.233 | 0.396 | 0.724 | — | — | — | — | |
| SMN2020.04 | 0.529 | 0.569 | 0.397 | — | — | — | 0.233 | 0.396 | 0.724 | — | |
| SMNDomain-specific post-training=false2020.09 | 0.529 | 0.569 | 0.397 | — | — | — | 0.233 | 0.396 | 0.724 | — | |
| SMN2020.12 | 0.529 | 0.569 | 0.397 | — | — | — | 0.233 | 0.396 | 0.724 | — | |
| SMNModel Type=non PrLM-based2022.03 | 0.529 | 0.569 | 0.397 | — | — | — | 0.233 | 0.396 | 0.724 | 0.926 | |
| Multi-View2020.04 | 0.505 | 0.543 | 0.342 | — | — | — | 0.202 | 0.35 | 0.729 | — | |
| Multi-ViewDomain-specific post-training=false2020.09 | 0.505 | 0.543 | 0.342 | — | — | — | 0.202 | 0.35 | 0.729 | — | |
| Multi-View2020.12 | 0.505 | 0.543 | 0.342 | — | — | — | 0.202 | 0.35 | 0.729 | — | |
| Match-LSTM2020.04 | 0.5 | 0.537 | 0.345 | — | — | — | 0.202 | 0.348 | 0.72 | — | |
| Match-LSTMDomain-specific post-training=false2020.09 | 0.5 | 0.537 | 0.345 | — | — | — | 0.202 | 0.348 | 0.72 | — | |
| Match-LSTM2020.12 | 0.5 | 0.537 | 0.345 | — | — | — | 0.202 | 0.348 | 0.72 | — | |
| MV-LSTM2020.04 | 0.498 | 0.538 | 0.348 | — | — | — | 0.202 | 0.351 | 0.71 | — | |
| MV-LSTMDomain-specific post-training=false2020.09 | 0.498 | 0.538 | 0.348 | — | — | — | 0.202 | 0.351 | 0.71 | — | |
| MV-LSTM2020.12 | 0.498 | 0.538 | 0.348 | — | — | — | 0.202 | 0.351 | 0.71 | — | |
| DL2R2020.04 | 0.488 | 0.527 | 0.33 | — | — | — | 0.193 | 0.342 | 0.705 | — | |
| DL2RDomain-specific post-training=false2020.09 | 0.488 | 0.527 | 0.33 | — | — | — | 0.193 | 0.342 | 0.705 | — | |
| DL2R2020.12 | 0.488 | 0.527 | 0.33 | — | — | — | 0.193 | 0.342 | 0.705 | — | |
| LSTM2020.04 | 0.485 | 0.527 | 0.32 | — | — | — | 0.187 | 0.343 | 0.72 | — | |
| LSTMDomain-specific post-training=false2020.09 | 0.485 | 0.537 | 0.32 | — | — | — | 0.187 | 0.343 | 0.72 | — | |
| LSTM2020.12 | 0.485 | 0.527 | 0.32 | — | — | — | 0.187 | 0.343 | 0.72 | — | |
| LSTMModel Type=non PrLM-based2022.03 | 0.485 | 0.537 | 0.32 | — | — | — | 0.187 | 0.343 | 0.72 | 0.638 | |
| BiLSTM2020.04 | 0.479 | 0.514 | 0.313 | — | — | — | 0.184 | 0.33 | 0.716 | — | |
| BiLSTMDomain-specific post-training=false2020.09 | 0.479 | 0.514 | 0.313 | — | — | — | 0.184 | 0.33 | 0.716 | — | |
| Bi-LSTM2020.12 | 0.479 | 0.514 | 0.313 | — | — | — | 0.184 | 0.33 | 0.716 | — | |
| CNN2020.04 | 0.417 | 0.44 | 0.226 | — | — | — | 0.121 | 0.252 | 0.647 | — | |
| CNNDomain-specific post-training=false2020.09 | 0.417 | 0.44 | 0.226 | — | — | — | 0.121 | 0.252 | 0.647 | — | |
| CNN2020.12 | 0.417 | 0.44 | 0.226 | — | — | — | 0.121 | 0.252 | 0.647 | — | |
| RNN2020.04 | 0.39 | 0.422 | 0.208 | — | — | — | 0.118 | 0.223 | 0.589 | — | |
| RNN2020.12 | 0.39 | 0.422 | 0.208 | — | — | — | 0.118 | 0.223 | 0.589 | — |