Poly-encoders: Transformer Architectures and Pre-training Strategies for Fast and Accurate Multi-sentence Scoring

Benchmark Model Rank Results
conversational-response-selection-on-douban-1Poly-encoder#9MAP: 0.608MRR: 0.650P@1: 0.475R10@1: 0.299R10@2: 0.494
conversational-response-selection-on-rrs-1Poly-encoder#1NDCG@3: 0.679NDCG@5: 0.765
conversational-response-selection-on-ubuntu-1Poly-encoder#6R10@1: 0.882R10@2: 0.949R10@5: 0.990