pith. sign in

arxiv: 2007.00808 · v2 · pith:GEYPE2F5new · submitted 2020-07-01 · 💻 cs.IR · cs.CL· cs.LG

Approximate Nearest Neighbor Negative Contrastive Learning for Dense Text Retrieval

classification 💻 cs.IR cs.CLcs.LG
keywords retrievaltrainingdensenegativeapproximatenearestneighborsparse
0
0 comments X
read the original abstract

Conducting text retrieval in a dense learned representation space has many intriguing advantages over sparse retrieval. Yet the effectiveness of dense retrieval (DR) often requires combination with sparse retrieval. In this paper, we identify that the main bottleneck is in the training mechanisms, where the negative instances used in training are not representative of the irrelevant documents in testing. This paper presents Approximate nearest neighbor Negative Contrastive Estimation (ANCE), a training mechanism that constructs negatives from an Approximate Nearest Neighbor (ANN) index of the corpus, which is parallelly updated with the learning process to select more realistic negative training instances. This fundamentally resolves the discrepancy between the data distribution used in the training and testing of DR. In our experiments, ANCE boosts the BERT-Siamese DR model to outperform all competitive dense and sparse retrieval baselines. It nearly matches the accuracy of sparse-retrieval-and-BERT-reranking using dot-product in the ANCE-learned representation space and provides almost 100x speed-up.

This paper has not been read by Pith yet.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 34 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Dense Passage Retrieval for Open-Domain Question Answering

    cs.CL 2020-04 accept novelty 8.0

    Dense dual-encoder retrievers outperform BM25 by 9-19% absolute in top-20 passage retrieval accuracy across open-domain QA datasets and enable new state-of-the-art end-to-end QA results.

  2. Generative Retrieval for Table Union Search

    cs.DB 2026-07 unverdicted novelty 7.0

    GenTUS reformulates table union search as constrained generative retrieval over semantic table identifiers, achieving top retrieval quality on seven benchmarks with lower latency and storage costs.

  3. MMed-Bench-IR: A Heterogeneous Benchmark for Multilingual Medical Information Retrieval

    cs.CL 2026-06 unverdicted novelty 7.0

    MMed-Bench-IR is a new heterogeneous benchmark spanning 6 languages and three non-overlapping tasks that exposes severe cross-lingual drops in biomedical retrieval performance.

  4. Can LLM Rerankers Predict Their Own Ranking Performance?

    cs.IR 2026-06 unverdicted novelty 7.0

    LLM rerankers can internally predict ranking quality via self-consistency of sampled outputs, matching SOTA external QPP while direct confidence is overconfident; supervised token-efficient methods improve calibration.

  5. Test-Time Training for Zero-Resource Dense Retrieval Reranking

    cs.IR 2026-05 unverdicted novelty 7.0

    DART adapts a scoring matrix at inference time via gradient updates on pseudo-labels from top/bottom documents to gain +2.1% mean NDCG@10 on six BEIR benchmarks with under 10ms added latency.

  6. BOOKMARKS: Efficient Active Storyline Memory for Role-playing

    cs.CL 2026-05 unverdicted novelty 7.0

    BOOKMARKS introduces searchable bookmarks as reusable answers to storyline questions, enabling active initialization and passive synchronization for more consistent role-playing agent memory than recurrent summarization.

  7. ConRetroBert: EMA Stabilized Dual Encoders for Template-Based Single-Step Retrosynthesis

    cs.LG 2026-05 unverdicted novelty 7.0

    ConRetroBert achieves 62.4% top-1 accuracy on USPTO-50k by combining contrastive pretraining, hard-negative listwise ranking, and EMA-stabilized dual encoders for template retrieval in retrosynthesis.

  8. Latent Abstraction for Retrieval-Augmented Generation

    cs.CL 2026-04 unverdicted novelty 7.0

    LAnR unifies retrieval-augmented generation inside a single LLM by deriving dense retrieval vectors from a [PRED] token's hidden states and using entropy to adaptively stop retrieval, outperforming prior RAG on six QA...

  9. VLM2Vec: Training Vision-Language Models for Massive Multimodal Embedding Tasks

    cs.CV 2024-10 conditional novelty 7.0

    VLM2Vec converts state-of-the-art vision-language models into universal multimodal embedders via contrastive training on the new MMEB benchmark, delivering 10-20% absolute gains over prior models on both in-distributi...

  10. M3-Embedding: Multi-Linguality, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation

    cs.CL 2024-02 unverdicted novelty 7.0

    M3-Embedding is a single model for multi-lingual, multi-functional, and multi-granular text embeddings trained via self-knowledge distillation that achieves new state-of-the-art results on multilingual, cross-lingual,...

  11. C-Pack: Packed Resources For General Chinese Embeddings

    cs.CL 2023-09 accept novelty 7.0

    C-Pack releases a new Chinese embedding benchmark, large training dataset, and optimized models that outperform priors by up to 10% on C-MTEB while also delivering English SOTA results.

  12. BitNet Text Embeddings

    cs.CL 2026-06 unverdicted novelty 6.0

    BITEMBED converts LLM backbones to ternary BitNet-style encoders, adapts them with contrastive pre-training and teacher distillation, and produces text embeddings at multiple precisions that perform comparably to full...

  13. Conflict-Aware Retriever Editing for Knowledge Injection Attacks on LLM-Based RAG Systems

    cs.CR 2026-06 unverdicted novelty 6.0

    CAREATTACK adapts closed-form parameter editing with graph-based conflict resolution and lightweight anchor repair to promote malicious passages in RAG retrieval while limiting side effects on non-target queries.

  14. CoDeR: Local Constraint-Compatible Retrieval Beyond Semantic Similarity

    cs.IR 2026-06 unverdicted novelty 6.0

    CoDeR augments standard topical dense retrieval with a bi-encoder compatibility scorer trained via contrastive lexical-polarity supervision to reduce early exposure to constraint-violating documents.

  15. Beyond Topical Similarity: Contrastive Evidence Retrieval with Interpretable Attention Alignment in RAG

    cs.CL 2026-05 unverdicted novelty 6.0

    CERA fine-tunes a dense retriever with triplet contrastive learning plus attention alignment to human rationales, claiming better retrieval effectiveness and faithfulness on clinical trial reports than Contriever and ...

  16. SemBridge: Language Transfer in Sparse Encoders via Multilingual Semantic Bridges

    cs.IR 2026-05 unverdicted novelty 6.0

    SemBridge initializes target tokens in sparse encoders via linear combinations of semantically related source tokens selected through multilingual dense bridge embeddings, yielding better zero-shot and fine-tuned retr...

  17. Why Advanced Encoders Lag on Sparse Retrieval? The Answer and an Approach to Bridging Vocabulary Gaps

    cs.IR 2026-04 conditional novelty 6.0

    Transferring modern encoders to normalized (lowercased) vocabularies via geometric embedding initialization and activation calibration closes the performance gap in learned sparse retrieval, achieving 52.4 nDCG on BEIR.

  18. Learning Chain Of Thoughts Prompts for Predicting Entities, Relations, and even Literals on Knowledge Graphs

    cs.CL 2026-04 unverdicted novelty 6.0

    RALP learns string-based chain-of-thought prompts as scoring functions for knowledge graph triples using Bayesian optimization from fewer than 30 examples, improving link prediction MRR by over 5% and achieving over 8...

  19. ARHN: Answer-Centric Relabeling of Hard Negatives with Open-Source LLMs for Dense Retrieval

    cs.IR 2026-04 unverdicted novelty 6.0

    ARHN refines hard-negative training data for dense retrieval by using LLMs to convert answer-containing passages into additional positives and exclude answer-containing passages from the negative set.

  20. Data, Not Model: Explaining Bias toward LLM Texts in Neural Retrievers

    cs.IR 2026-04 unverdicted novelty 6.0

    Bias toward LLM texts in neural retrievers arises from artifact imbalances between positive and negative documents in training data that are absorbed during contrastive learning.

  21. Atlas: Few-shot Learning with Retrieval Augmented Language Models

    cs.CL 2022-08 unverdicted novelty 6.0

    Atlas reaches over 42% accuracy on Natural Questions with only 64 examples, outperforming a 540B-parameter model by 3% with 50x fewer parameters.

  22. Text and Code Embeddings by Contrastive Pre-Training

    cs.CL 2022-01 unverdicted novelty 6.0

    Contrastive pre-training on unsupervised data at scale creates text and code embeddings that set new state-of-the-art results on classification and semantic search benchmarks.

  23. Unsupervised Dense Information Retrieval with Contrastive Learning

    cs.IR 2021-12 unverdicted novelty 6.0

    Contrastive learning trains unsupervised dense retrievers that beat BM25 on most BEIR datasets and support cross-lingual retrieval across scripts.

  24. Subtraction Gets You More: Gap-Aware Retrieval for Multimodal Multi-Hop QA

    cs.IR 2026-05 unverdicted novelty 5.0

    GRAIL uses context-subtractive query steering at the embedding level to break semantic anchoring in iterative retrieval, achieving a 40.3% macro-averaged gain on MultimodalQA via a hybrid routing framework.

  25. PRA-RAG: Provably Robust Aggregation in Retrieval-Augmented Generation against Retrieval Corruption

    cs.IR 2026-05 unverdicted novelty 5.0

    PRA-RAG is a new aggregation algorithm for RAG that claims provable robustness bounds against poisoned retrieved texts and reduces attack success rate to 1% while keeping 71% accuracy.

  26. The Role of Vocabularies in Learning Sparse Representations for Ranking

    cs.IR 2025-09 unverdicted novelty 5.0

    Larger 100K vocabularies in SPLADE models, especially those initialized with ESPLADE pretraining, improve retrieval effectiveness after pruning compared to 32K baselines while keeping similar efficiency.

  27. Improving Korean-English Cross-Lingual Retrieval: A Data-Centric Study of Language Composition and Model Merging

    cs.IR 2025-07 unverdicted novelty 5.0

    Language composition in training data creates opposing effects on CLIR and mono-IR performance for Korean-English retrieval, which model merging can partially resolve.

  28. Don't Retrieve, Generate: Prompting LLMs for Synthetic Training Data in Dense Retrieval

    cs.IR 2025-04 unverdicted novelty 5.0

    LLM-generated synthetic hard negatives for training dense retrievers consistently underperform corpus-mined negatives from BM25 and cross-encoders across 10 BEIR datasets, with non-monotonic gains from scaling the gen...

  29. Decoupling Semantics and Logic: A Training-Free Coarse-to-Fine Pipeline for Video Retrieval-Augmented Generation

    cs.CV 2026-06 unverdicted novelty 4.0

    A cascaded training-free Video RAG pipeline decouples high-recall semantic prefetching from LLM-driven logical reranking to improve precision on cross-lingual long-video tasks with persona constraints.

  30. Granite Embedding Multilingual R2 Models

    cs.IR 2026-05 unverdicted novelty 4.0

    Granite Embedding Multilingual R2 releases 311M and 97M parameter bi-encoder models that achieve state-of-the-art retrieval performance on multilingual text, code, long-document, and reasoning datasets.

  31. CogRAG: Tackling Heterogeneous Cognitive Demands in RAG via Stratified Retrieval and Reasoning

    cs.CL 2026-04 unverdicted novelty 4.0

    CogRAG+ raises LLM accuracy on a dietitian exam to 85.8% by using dual-path retrieval and structured reasoning templates.

  32. Position: LLM Watermarking Should Align Stakeholders' Incentives for Practical Adoption

    cs.CR 2025-10 unverdicted novelty 4.0

    LLM watermarking adoption is limited by misaligned stakeholder incentives; incentive-aligned approaches such as in-context watermarking can enable practical use in targeted domains like education and peer review.

  33. Query pipeline optimization for cancer patient question answering systems

    cs.CL 2024-12 unverdicted novelty 4.0

    Three-aspect RAG query pipeline optimization for cancer patient QA introduces HSRDR and SEOS and reports 5.24% accuracy gain on Claude-3-haiku versus chain-of-thought on a custom dataset.

  34. A Survey of Scaling in Large Language Model Reasoning

    cs.AI 2025-04 unverdicted novelty 3.0

    A survey categorizing scaling in LLM reasoning across input size, steps, rounds, training, and future directions, noting that scaling can negatively affect performance.