Do Neural Retrievers Prefer Certain Documents? Evidence of Learned Relevance Priors

• URL: http://arxiv.org/abs/2606.02814v1
• Date: Mon, 01 Jun 2026 19:31:28 GMT
• Title: Do Neural Retrievers Prefer Certain Documents? Evidence of Learned Relevance Priors
• Abstract summary: We investigate whether supervised bi-encoder retrievers implicitly learn a document-level relevance prior.<n>We find that supervised neural retrievers encode relevance priors that generalize to unseen documents and are consistent across models.<n>We find that judged-relevant documents tend to be comprehensive, self-contained summaries of mainstream topics, while niche, fragmentary, or highly technical content is often left unjudged.
• Score: 2.5681874692250193
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Neural retrievers are trained to estimate query-document relevance from annotated query-document pairs. Yet annotation protocols may not purely reflect relevance: they select only a subset of documents for labeling, and this selection can favor certain document types over others. We investigate whether supervised bi-encoder retrievers implicitly learn a document-level relevance prior: a query-independent signal encoded in their representation space as a side effect of training on annotated data. We estimate this prior by training simple classifiers on frozen document embeddings and evaluate three state-of-the-art retrievers across multiple IR benchmarks. We find that supervised neural retrievers encode relevance priors that generalize to unseen documents and are consistent across models. These priors create a findability gap: documents with lower prior are systematically harder to retrieve, even when genuinely relevant. This effect appears in supervised dense retrievers but is weaker and less consistent in BM25, and it persists under controlled matched-document comparisons. Using LLM-based explanations, we find that judged-relevant documents tend to be comprehensive, self-contained summaries of mainstream topics, while niche, fragmentary, or highly technical content is often left unjudged. Retrievers internalize this bias, ranking documents with these favored features higher than documents that lack them, independently of their actual relevance. Our findings expose a structural limitation of supervised retrieval: models trained on annotated data do not just learn relevance, but also the implicit document preferences in their training data.

Related papers

• ReLeVAnT: Relevance Lexical Vectors for Accurate Legal Text Classification [5.530730417910086]
  ReLeVAnT is a framework for legal document binary classification.<n>It uses n-gram processing, contrastive score matching, and a shallow neural network as the primary drivers for discriminative classification.<n>It has 99.3% accuracy and 98.7% F1 score on the LexGLUE dataset.
  arXiv  Detail & Related papers  (2026-04-24T07:13:24Z)
• Dual-View Training for Instruction-Following Information Retrieval [9.741583305127994]
  We propose a dual-view data synthesis strategy based on polarity reversal.<n>We show that our method improves performance on the FollowIR benchmark by 45%.
  arXiv  Detail & Related papers  (2026-04-20T21:10:29Z)
• Query Decomposition for RAG: Balancing Exploration-Exploitation [83.79639293409802]
  RAG systems address complex user requests by decomposing them into subqueries, retrieving potentially relevant documents for each, and then aggregating them to generate an answer.<n>We formulate query decomposition and document retrieval in an exploitation-exploration setting, where retrieving one document at a time builds a belief about the utility of a given sub-queries.<n>Our main finding is that estimating document relevance using rank information and human judgments yields a 35% gain in document-level precision, 15% increase in alpha-nDCG, and better performance on the downstream task of long-form generation.
  arXiv  Detail & Related papers  (2025-10-21T13:37:11Z)
• Beyond Contrastive Learning: Synthetic Data Enables List-wise Training with Multiple Levels of Relevance [30.879299174443812]
  In this work, we forgo real documents and annotations and use large language models to generate synthetic documents.<n>Our experiments on MS MARCO and BEIR benchmark show that our proposed approach outperforms conventional training with InfoNCE by a large margin.
  arXiv  Detail & Related papers  (2025-03-29T22:33:22Z)
• Collapse of Dense Retrievers: Short, Early, and Literal Biases Outranking Factual Evidence [56.09494651178128]
  Retrieval models are commonly used in Information Retrieval (IR) applications, such as Retrieval-Augmented Generation (RAG)<n>We quantify the impact of biases, such as a preference for shorter documents, on retrievers like Dragon+ and Contriever.<n>We uncover major vulnerabilities, showing retrievers favor shorter documents, early positions, repeated entities, and literal matches, all while ignoring the answer's presence!
  arXiv  Detail & Related papers  (2025-03-06T23:23:13Z)
• Learning Refined Document Representations for Dense Retrieval via Deliberate Thinking [58.69615583599489]
  Deliberate Thinking based Retriever (Debater) is a novel approach that enhances document representations by incorporating a step-by-step thinking process.<n>Debater significantly outperforms existing methods across several retrieval benchmarks.
  arXiv  Detail & Related papers  (2025-02-18T15:56:34Z)
• DAPR: A Benchmark on Document-Aware Passage Retrieval [57.45793782107218]
  We propose and name this task emphDocument-Aware Passage Retrieval (DAPR) While analyzing the errors of the State-of-The-Art (SoTA) passage retrievers, we find the major errors (53.5%) are due to missing document context. Our created benchmark enables future research on developing and comparing retrieval systems for the new task.
  arXiv  Detail & Related papers  (2023-05-23T10:39:57Z)
• Precise Zero-Shot Dense Retrieval without Relevance Labels [60.457378374671656]
  Hypothetical Document Embeddings(HyDE) is a zero-shot dense retrieval system. We show that HyDE significantly outperforms the state-of-the-art unsupervised dense retriever Contriever.
  arXiv  Detail & Related papers  (2022-12-20T18:09:52Z)
• CAPSTONE: Curriculum Sampling for Dense Retrieval with Document Expansion [68.19934563919192]
  We propose a curriculum sampling strategy that utilizes pseudo queries during training and progressively enhances the relevance between the generated query and the real query. Experimental results on both in-domain and out-of-domain datasets demonstrate that our approach outperforms previous dense retrieval models.
  arXiv  Detail & Related papers  (2022-12-18T15:57:46Z)
• Natural Logic-guided Autoregressive Multi-hop Document Retrieval for Fact Verification [21.04611844009438]
  We propose a novel retrieve-and-rerank method for multi-hop retrieval. It consists of a retriever that jointly scores documents in the knowledge source and sentences from previously retrieved documents. It is guided by a proof system that dynamically terminates the retrieval process if the evidence is deemed sufficient.
  arXiv  Detail & Related papers  (2022-12-10T11:32:38Z)
• Improving Query Representations for Dense Retrieval with Pseudo Relevance Feedback [29.719150565643965]
  This paper proposes ANCE-PRF, a new query encoder that uses pseudo relevance feedback (PRF) to improve query representations for dense retrieval. ANCE-PRF uses a BERT encoder that consumes the query and the top retrieved documents from a dense retrieval model, ANCE, and it learns to produce better query embeddings directly from relevance labels. Analysis shows that the PRF encoder effectively captures the relevant and complementary information from PRF documents, while ignoring the noise with its learned attention mechanism.
  arXiv  Detail & Related papers  (2021-08-30T18:10:26Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.