Pith. sign in

REVIEW 7 cited by

How to Train Your DRAGON: Diverse Augmentation Towards Generalizable Dense Retrieval

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2302.07452 v1 pith:7EZAARD7 submitted 2023-02-15 cs.IR cs.CL

How to Train Your DRAGON: Diverse Augmentation Towards Generalizable Dense Retrieval

classification cs.IR cs.CL
keywords augmentationretrievaldensediversegeneralizablesupervisedzero-shotachieve
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Various techniques have been developed in recent years to improve dense retrieval (DR), such as unsupervised contrastive learning and pseudo-query generation. Existing DRs, however, often suffer from effectiveness tradeoffs between supervised and zero-shot retrieval, which some argue was due to the limited model capacity. We contradict this hypothesis and show that a generalizable DR can be trained to achieve high accuracy in both supervised and zero-shot retrieval without increasing model size. In particular, we systematically examine the contrastive learning of DRs, under the framework of Data Augmentation (DA). Our study shows that common DA practices such as query augmentation with generative models and pseudo-relevance label creation using a cross-encoder, are often inefficient and sub-optimal. We hence propose a new DA approach with diverse queries and sources of supervision to progressively train a generalizable DR. As a result, DRAGON, our dense retriever trained with diverse augmentation, is the first BERT-base-sized DR to achieve state-of-the-art effectiveness in both supervised and zero-shot evaluations and even competes with models using more complex late interaction (ColBERTv2 and SPLADE++).

discussion (0)

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

Forward citations

Cited by 7 Pith papers

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

  1. Mutual Linearity in and out of Stationarity for Markov Jump Processes: A Trajectory-Based Approach

    cond-mat.stat-mech 2026-04 unverdicted novelty 7.0

    A trajectory-level derivation shows mutual linearity holds for non-stationary Markov jump processes and generalizes to other systems.

  2. From Standalone LLMs to Integrated Intelligence: A Survey of Compound Al Systems

    cs.MA 2025-06 accept novelty 7.0

    A survey that defines Compound AI Systems, proposes a multi-dimensional taxonomy based on component roles and orchestration strategies, reviews four foundational paradigms, and identifies key challenges for future research.

  3. KAMR: Grounding Generation via Knowledge-Aligned Multi-hop Retrieval

    cs.IR 2026-07 conditional novelty 6.0

    Partial-alignment contrastive pretraining plus anchor-then-expand graph retrieval improves multi-hop KG evidence recovery and downstream QA over strong dense and graph RAG baselines.

  4. AV-SQL: Decomposing Complex Text-to-SQL Queries with Agentic Views

    cs.DB 2026-04 unverdicted novelty 6.0

    AV-SQL uses a pipeline of LLM agents to generate intermediate CTE views that decompose complex Text-to-SQL queries, reaching 70.38% execution accuracy on Spider 2.0.

  5. 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.

  6. Mutual Linearity in and out of Stationarity for Markov Jump Processes: A Trajectory-Based Approach

    cond-mat.stat-mech 2026-04 unverdicted novelty 5.0

    Trajectory-level linear response yields mutual linearity of observables under single-edge rate perturbation for Markov jump processes, including non-stationary state and counting observables.

  7. Beyond Hard Negatives: The Importance of Score Distribution in Knowledge Distillation for Dense Retrieval

    cs.IR 2026-04 unverdicted novelty 5.0

    Stratified sampling preserving teacher score distribution outperforms hard-negative mining as a robust baseline for knowledge distillation in dense retrieval.