Efficient Dynamic Hard Negative Sampling for Dialogue Selection

Benchmark Model Rank Results
conversational-response-selection-on-eBERT-FP+EDHNS#1R10@1: 0.957R10@2: 0.986R10@5: 0.997
conversational-response-selection-on-ubuntu-1BERT-FP+EDHNS#2R10@1: 0.917R10@2: 0.965R10@5: 0.994