Papers › Poly-encoders: Transformer Architectures and Pre-training Strategies for Fast and...

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

22 Apr 2019arXiv:1905.01969archive 2025-07-28

Samuel Humeau, Kurt Shuster, Marie-Anne Lachaux, Jason Weston

The use of deep pre-trained bidirectional transformers has led to remarkable progress in a number of applications (Devlin et al., 2018). For tasks that make pairwise comparisons between sequences, matching a given input with a corresponding label, two approaches are common: Cross-encoders performing full self-attention over the pair and Bi-encoders encoding the pair separately. The former often performs better, but is too slow for practical use. In this work, we develop a new transformer architecture, the Poly-encoder, that learns global rather than token level self-attention features. We perform a detailed comparison of all three approaches, including what pre-training and fine-tuning strategies work best. We show our models achieve state-of-the-art results on three existing tasks; that Poly-encoders are faster than Cross-encoders and more accurate than Bi-encoders; and that the best results are obtained by pre-training on large datasets similar to the downstream tasks.

PaperPDFCode

In Syntology View this paper on Syntology, its page in Syntology's graph. That page lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

Alexey-Borisov/3_course_diary mentioned on GitHubSyntology: not harvested report
chijames/Poly-Encoder mentioned on GitHubpytorchMITSyntology: not harvested report
csong27/collision-bert mentioned on GitHubpytorchSyntology: not harvested report
fangrouli/Document-embedding-generation-models mentioned on GitHubpytorchSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks (arXiv:1908.10084; that paper's own run record: 33 ran (of which 9 constructed an object rather than computing a result; 30 with no instrument failure: 1 honoured, 0 violated, 29 with no contract checked; 3 where Syntology's instrument failed) · 25 unverified); nothing checks that they implement this paper's method. report
i2r-simmc/i2r-simmc-2020 mentioned on GitHubpytorchMITSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation,... (arXiv:1910.13461; that paper's own run record: 40 ran (of which 7 constructed an object rather than computing a result; 30 with no instrument failure: 1 honoured, 2 violated, 27 with no contract checked; 10 where Syntology's instrument failed) · 13 unverified); nothing checks that they implement this paper's method. report
llStringll/Poly-encoders mentioned on GitHubpytorchSyntology: not harvested report
sfzhou5678/PolyEncoder mentioned on GitHubpytorchSyntology: not harvested report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

A paper named beside a repository with no sample linked to this paper is shown with that paper's own run record, not this paper's: “ran” means executed on a synthesized input, not that the code is correct, and “where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Papers are listed in arXiv-id order, at most three per repository.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Conversational Response SelectionSentence

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Conversational Response Selection DSTC7 Ubuntu Bi-encoder (v2) 1-of-100 Accuracy 70.9% #2 of 5 Archive leaderboard report
Conversational Response Selection DSTC7 Ubuntu Bi-encoder 1-of-100 Accuracy 66.3% #3 of 5 Archive leaderboard report
Conversational Response Selection Douban Poly-encoder MAP 0.608 #9 of 16 Archive leaderboard report
Conversational Response Selection Douban Poly-encoder MRR 0.650 #9 of 16 Archive leaderboard report
Conversational Response Selection Douban Poly-encoder P@1 0.475 #9 of 16 Archive leaderboard report
Conversational Response Selection Douban Poly-encoder R10@1 0.299 #9 of 16 Archive leaderboard report
Conversational Response Selection Douban Poly-encoder R10@2 0.494 #9 of 16 Archive leaderboard report
Conversational Response Selection Douban Poly-encoder R10@5 0.822 #9 of 16 Archive leaderboard report
Conversational Response Selection RRS Ranking Test Poly-encoder NDCG@3 0.679 #1 of 4 Archive leaderboard report
Conversational Response Selection RRS Ranking Test Poly-encoder NDCG@5 0.765 #1 of 4 Archive leaderboard report
Conversational Response Selection Ubuntu Dialogue (v1, Ranking) Poly-encoder R10@1 0.882 #8 of 25 Archive leaderboard report
Conversational Response Selection Ubuntu Dialogue (v1, Ranking) Poly-encoder R10@2 0.949 #8 of 25 Archive leaderboard report
Conversational Response Selection Ubuntu Dialogue (v1, Ranking) Poly-encoder R10@5 0.990 #8 of 25 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Absolute Position EncodingsAdamAttentionBPEDense ConnectionsDropoutLabel SmoothingLayer NormalizationLinear LayerMulti-Head AttentionPosition-Wise Feed-Forward LayerReLUResidual ConnectionSoftmaxTransformer

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections