AI Paper Atlas

Research map / Vision-language and multimodal / CLIP and zero-shot

LexMAE: Lexicon-Bottlenecked Pretraining for Large-Scale Retrieval

Tao Shen, Xiubo Geng, Chongyang Tao, Can Xu, Xiaolong Huang, Binxing Jiao, Linjun Yang, Daxin Jiang

ICLR 2023 · Vision-language and multimodal › CLIP and zero-shot · approach: Language models for zero-shot reasoning · 13 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Pretrain with lexicon-bottlenecked masked autoencoding rather than the strongest baseline initializationLexicon-weighting retrieval on MS-Marco Dev passage · MS-Marco Dev passageMRR@10: 41.1 -> 42.6Improved
Pretrain with LexMAE rather than coCondenser initializationLexicon-weighting retrieval on MS-Marco Dev passage · MS-Marco Dev passageMRR@10: +5.1% MRR@10Improved
Use LexMAE for zero-shot transferBEIR retrieval · BEIR zero-shot average (12 datasets)nDCG@10: 47.0 -> 48.7Improved
Keep the CBoW bottleneck rather than disabling it (ablation)LexMAE pretraining for MS-Marco Dev passage retrieval · MS-Marco Dev passage, first-stage fine-tuningMRR@10: 38.5 -> 39.3Improved
Use a CBoW bottleneck rather than a dense [CLS] bottleneck (ablation)LexMAE pretraining for MS-Marco Dev passage retrieval · MS-Marco Dev passage, first-stage fine-tuningMRR@10: 38.6 -> 39.3Improved
Use softmax normalization rather than saturated norm in the CBoW bottleneck (ablation)LexMAE pretraining for MS-Marco Dev passage retrieval · MS-Marco Dev passage, first-stage fine-tuningMRR@10: 39.2 -> 39.3Improved
Stop gradients from flowing to the bottleneck word embeddings (ablation)LexMAE pretraining for MS-Marco Dev passage retrieval · MS-Marco Dev passage, first-stage fine-tuningMRR@10: 38.9 -> 39.3Improved
Keep encoder and decoder LM heads unshared (ablation)LexMAE pretraining for MS-Marco Dev passage retrieval · MS-Marco Dev passage, first-stage fine-tuningMRR@10: 39.1 -> 39.3Improved
Do not add a separate LM head for bottleneck logits (ablation)LexMAE pretraining for MS-Marco Dev passage retrieval · MS-Marco Dev passage, first-stage fine-tuningMRR@10: 39.2 -> 39.3Improved
Use BERT rather than DistilBERT initialization (ablation)LexMAE pretraining for MS-Marco Dev passage retrieval · MS-Marco Dev passage, first-stage fine-tuningMRR@10: 38.4 -> 39.3Improved
Use the inclusive masking strategy rather than exclusive masking (ablation)LexMAE pretraining for MS-Marco Dev passage retrieval · MS-Marco Dev passage, first-stage fine-tuningMRR@10: 39.1 -> 39.3Improved
Use inclusive rather than fully random masking (ablation)LexMAE pretraining for MS-Marco Dev passage retrieval · MS-Marco Dev passage, first-stage fine-tuningMRR@10: 39.2 -> 39.3No clear effect

Later papers that compared with this method

1 later accepted paper reports LexMAE as a baseline in their results tables; 0 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy