Overview

Dense retrieval encodes queries and candidates as vectors, then uses nearest-neighbor search to find relevant results, often where exact keyword matching is inadequate. Common approaches include dual encoders trained with contrastive learning and hard negatives, late-interaction models that match token-level representations, and hybrid systems that combine dense scores with lexical retrieval or reranking. Researchers evaluate these methods on benchmarks such as BEIR and domain-specific tests, while also developing compact indexes, multilingual and specialized retrievers, and retrieval for long documents or agent tasks. A central tension is that stronger models and richer representations can improve difficult or compositional search but cost more to train and serve; generalization, robustness, fairness, and reliable constraint handling remain open challenges.

Latest papers

160 papers15 papers in the last four weeks, up 7% on the four weeks before. 0.1% of all new papers.

Sep 11, 2026

Mixture-of-Experts Language Models Can Be Strong and Efficient Retrievers

Recent work has shown that fine-tuning decoder-only large language models (LLMs) for retrieval yields strong first-stage retrievers, with effectiveness improving as backbones grow in size. However, every query and document must pass through the full model, so encoding cost increases with model size. Mixture-of-Experts (MoE) LLMs activate only a subset of parameters per token and are widely used to scale generative models, yet remain underexplored as retrievers. We systematically study MoE backbones for retrieval by training MoE and dense LLMs from several families using the same procedure, evaluating them across diverse datasets, and measuring query encoding time under the same serving configuration. We show that MoE retrievers outperform dense retrievers with comparable active parameter counts by up to 3.0 nDCG@10 points on BEIR. One of our strongest MoE retrievers matches an 8B dense retriever with 59% fewer active parameters and 18% lower query encoding time. We further show that the number of experts used for query encoding can be reduced without retraining or re-indexing, retaining more than 99% of retrieval effectiveness while reducing query encoding time by up to 26%. Recent rerankers provide only modest additional gains over strong MoE first stages, which often match or exceed the reranked configurations we evaluate. Together, these results show that MoE LLMs can be strong and efficient first-stage retrievers.
Sep 11, 2026

Parameter-Efficient Retrievers for Polish and European Languages

Dense retrieval systems increasingly rely on multi-billion-parameter language models, whose memory and computational requirements make large-scale indexing, frequent corpus updates, and low-latency serving costly. We present a three-stage training pipeline for developing compact and efficient retrievers that remain competitive with substantially larger models. The pipeline combines cross-lingual alignment, relational knowledge distillation, and contrastive fine-tuning. It requires no original ground-truth relevance labels, relying exclusively on supervision generated by strong embedding models and rerankers utilised as teachers. Using this pipeline, we develop PolDense and EuroDense, both supporting contexts of up to 8,192 tokens. PolDense is a family of six Polish retrievers ranging from 17M to 1B parameters. EuroDense is a 435M-parameter retriever supporting nine European languages. We conduct an extensive evaluation covering 41 Polish and 150 multilingual retrieval tasks. The results demonstrate strong quality-efficiency trade-offs. PolDense-1B outperforms the evaluated retrievers with up to 9B parameters, while the PolDense family forms the Pareto frontier across model sizes. Among the evaluated models below 1B parameters, EuroDense ranks first in both task-averaged and language-averaged performance and leads in seven of nine languages. We release all models publicly.
Sep 8, 2026

Reading a Legal Question Word by Word: Embedding Trajectories of 2,144 Vietnamese Legal Headlines

A dense retriever encodes a question as one vector, but the question arrives one word at a time. We read 2,144 held-out headlines from Thu Vien Phap Luat (Vietnamese legal library) word by word with Nemotron-3-Embed 8B/1B and Qwen3-Embedding 8B/0.6B, encoding 65,444 prefixes against 20,034 articles, plus every prefix of 3,438 sub-questions from 1,112 multi-question headlines and of 168 answers. (i) The gold article becomes rank 1 after a median of 6-7 content words in every encoder, before the interrogative frame is read, and stays there to the end in 78-85% of cases. (ii) In a multi-question headline the lock is inside the first sub-question 94-98% of the time; the second leaves rank unchanged in 89-95%; encoded alone, the second reaches rank 1 in 42-58% vs 91-96% for the first, at the same lock word (95-97% identical). (iii) Numbers, dates and instrument identifiers move the embedding twice as far as content words and four times as far as interrogative words; 72-78% of steps move toward the gold article, and the closing interrogative frame moves against that direction in 95-99% of headlines. (iv) Rank/cosine clustering yields six archetypes (instant, typical, unstable, late, never-locking) that differ by legal area and form (chi-squared p < 1e-8): real-estate and litigation headlines never lock on a number; environmental and accounting headlines do so a third of the time. (v) An answer read word by word retrieves its article after 8-16 words and addresses the sub-questions in order asked in 83-89% of cases. (vi) A word's step keeps a consistent direction across headlines (cosine 0.25-0.33; 0.44-0.60 for numbers); a preceding question rotates that step by about 60 degrees and a greeting by about 30 degrees; steps shrink as i^{-0.8}; and a two-question headline is within 12-17 degrees of a linear mix of its two questions. We call this a context-modulated additive walk.
Sep 8, 2026

OntologyBench: Can Dense Retrieval Satisfy Structured Biomedical Constraints?

We introduce OntologyBench, a tiered biomedical retrieval benchmark comprising 471,854 training and 125,744 evaluation query-document relevance pairs across concept grounding, relational retrieval, and compositional phenotype-based retrieval. Although these tasks can be tractable using ontology-aware reference methods, across task tiers, embedding performance is generally lower on relational and compositional tasks than on concept-grounding tasks. Fine-tuning on ontology-derived supervision improves performance on several relational and compositional tasks, whereas the evaluated reranking and LLM-based candidate-scoring methods provide little or no end-to-end improvement. Errors frequently reflect diseases matching only subsets of the phenotype evidence. These findings indicate that the evaluated embedding and reranking configurations do not reliably recover the compatibility encoded by the selected ontology relations and phenotype combinations and motivate retrieval systems that better integrate learned representations with structured biomedical knowledge.
Sep 7, 2026

EigenLI: Spectral Approximations to Late Interaction

Late-interaction models such as ColBERT achieve strong effectiveness by representing each document with many token-level vectors, but this expressivity leads to large indexing cost, storage footprints and expensive MaxSim scoring. We show that late-interaction representations exhibit an intrinsic low-rank structure: document token embeddings concentrate in a low-dimensional subspace that preserves most of the retrieval signal. Leveraging this observation, we introduce EigenLI, a spectral approximation framework that compresses late-interaction representations via document-specific low-dimensional subspaces. Unlike clustering or pooling methods, EigenLI identifies the dominant eigendirections of each document and uses them to construct reduced interaction representations. Empirically, kk-EigenLI with k≤32k \le 32 outperforms k-means and Ward clustering based pooling methods on ColBERTv2 and AnswerAI-ColBERT-small; GTE-ModernColBERT exhibits a different tradeoff at k=32k=32, where clustering methods perform better. The same spectral construction also yields EigenLI-SV, an ANN-compatible single-vector representation derived from the second-order summary of the reduced structure. Across multiple datasets and all three text models, EigenLI-SV consistently outperforms comparable single-vector surrogates such as MUVERA.
Sep 7, 2026

Matryoshka Hash Representations for Model-Aware Compact Semantic Retrieval

Retrieval-augmented generation (RAG) depends on dense retrieval: each document is stored as a learned vector, and queries are answered by nearest-neighbor search. Keeping one full-precision vector per document dominates index cost at corpus scale, so systems replace each vector with a short code of a few bytes (quantization). Standard quantizers such as product quantization (PQ) pick the code that best reconstructs the original vector, but reconstruction is only a proxy for retrieval: at a few bytes per document, a model-aware code trained on the retriever's ranking signal uses the budget better. A single code is even more useful if its short prefixes are each directly searchable, letting a deployment set its efficiency-quality operating point without re-encoding the corpus. But training all prefixes under one objective makes the early bits a compromise across budgets: short codes improve while the full-width code degrades. Binary codes sharpen the conflict, since a gradient step either flips a stored bit or leaves it unchanged. We introduce Matryoshka Hash Representations (MHR), a two-stage procedure that separates full-width training from prefix organization. MHR first learns a longer binary code, then freezes the model and trains zero-initialized residual code adaptors for directly searchable prefixes. Documents are stored at one bit per coordinate, queries keep continuous logits as in PQ, and search runs on FAISS FastScan. Trained on MS MARCO and zero-shot transferred to seven BEIR datasets, one MHR code reaches 0.502 NDCG@10 at 16 bytes and 0.556 at 32 bytes, against 0.404 and 0.524 for the best compact baselines; the gains hold across training seeds, on all 13 public BEIR datasets, and against off-the-shelf binary Matryoshka embeddings with stronger backbones. The same code also helps shortlist candidates for reranking and prune low-storage graph indexes such as LEANN.
Sep 5, 2026

ExpertLens: Visualizing Embedding Spaces for Post-Hoc Explainability in MoE Enhanced Retrievers

Neural models, including dense retrievers, have been widely adopted in Information Retrieval (IR), often delivering state-of-the-art performance. Despite their effectiveness, these models operate as black boxes, limiting the interpretability of their ranking decisions. Existing post-hoc explainability methods for neural rankers primarily focus on feature-level attributions, which can be insufficient to capture the complexity of learned embedding spaces. In this work, we propose ExpertLens, a post-hoc explainability framework for Mixture-of-Experts (MoE)-enhanced dense retrievers that shifts focus from local scalar feature importance to representation-level global interpretability. ExpertLens leverages discriminative embedding space visualizations jointly with automatically extracted Concept Activation Vectors to reveal how expert routing drives embedding space formulation and retrieval effectiveness. Experiments across five IR benchmarks and two MoE-enhanced dense retrievers show that expert routing consistently improves embedding space structure, positioning queries and their relevant documents into better-defined geometric neighborhoods. Analysis of expert subspaces further reveals general-purpose dominant experts, along with minority experts exhibiting distinct linguistic specialization, with subspaces arranged according to multi-semantic concept similarity. Our code is publicly available.
Sep 3, 2026

Comparing Retrieval Methods for Academic Advisor Discovery: A Six-Method Study of 768 CS Faculty Profiles Across 9 US Universities

We present a comparative evaluation of six information retrieval methods for the task of academic advisor discovery: ranking CS faculty members by relevance to a graduate applicant's research interest statement. The methods span sparse lexical matching (Jaccard overlap, TF-IDF, BM25), dense semantic retrieval (all-MiniLM-L6-v2 sentence embeddings), hybrid score fusion, and learning-to-rank. Evaluation uses a new domain-specific collection: 768 faculty profiles scraped from 9 US CS departments, with 162 graded relevance judgments (grade 0/1/2) across 5 queries representing distinct graduate student research profiles. Across all five queries, Reranked achieves the highest mean NDCG@10 (0.477, std 0.138), followed by Semantic (0.450), Hybrid (0.421), BM25 (0.406), Jaccard (0.303), and TF-IDF (0.246). After Bonferroni correction across all 15 pairwise comparisons, TF-IDF is significantly worse than BM25, Semantic, Hybrid, and Reranked; no other pairwise difference survives correction at 5 queries. A field ablation reveals that biography alone (NDCG 0.634) outperforms the full model combining biography with research area tags (0.593). A controlled experiment shows that concatenating arXiv paper abstracts reduces NDCG@10 by 0.176, motivating a late-fusion architecture. All code, scrapers, and relevance labels are released openly.
Sep 3, 2026

Spruce: Scalable Private Outsourced Retrieval Using Compact Embeddings

Retrieval-Augmented Generation (RAG) has made dense retrieval over large document collections a standard building block. Organizations increasingly outsource vector indexes to untrusted clouds, exposing proprietary corpora and user queries. Cryptographic protection is challenging because each query searches corpus-scale state, causing computation, correlated randomness, and communication to grow with the corpus. At million-document scale, a naive secure implementation takes minutes and about 90 GB of communication per query. Even recent optimized systems require 10--22 seconds. We propose Spruce (Scalable Private Outsourced Retrieval Using Compact Embeddings), which co-designs representations with the cryptographic protocol. Spruce learns compact binary codes that preserve candidates for full-precision reranking, replacing corpus-wide embedding scoring with efficient Hamming-distance computation under two-server multi-party computation (MPC). A corpus-calibrated fixed-radius protocol avoids multi-round candidate selection while preserving retrieval quality. Spruce also provides private cluster pruning, which trades minor quality loss for substantially less computation, and a one-core owner-operated dealer that removes cloud OT preprocessing bottlenecks. Across four corpora containing 383K--5.42M documents, Spruce preserves the original search quality with median candidate sets of only 382--1,952. At 10 Gbps inter-server bandwidth, full scans take 0.21--2.97 seconds, 4.84.8--6.7×6.7\times faster than the closest measured prior work. Private pruning takes 0.06--1.09 seconds, achieves 13.113.1--22.9×22.9\times speedups, and retains 93.9%93.9\%--97.3%97.3\% of full-float NDCG. On the largest corpus, pruning and the dealer jointly improve sustained throughput by 31.5×31.5\times at 1 Gbps per link.
Sep 1, 2026

Retrieved but not ranked: surface-form bias in structural retrieval, from mathematics to agent trajectories

We evaluate embedding retrieval where surface form and meaning are pulled apart on purpose: retrieving items that share underlying structure but not wording, in two unrelated domains under one protocol, competition mathematics (MathNet-Retrieve; 500 queries, 117,088-item corpus) and embodied-agent trajectories (ALFWorld-derived; 118 queries, 336 trajectories). In mathematics the failure is complete: strict Hit@1 at the heaviest disguise tier is 0.0% for both production embedders (bootstrap 95% CI [0.0, 0.0]) while the correct item sits in the top 10 nearly always, and in 95.2 to 99.8% of misses the winner is more lexically similar to the query than the correct answer. In trajectories, where surface variation is incidental, the same models land at or near hypergeometric chance when gold must involve a different object, and below chance for all three embedders once gold must differ in object and receptacle: retrieval anchors on literal tokens, not task structure. A lexical reranker control hurts in mathematics and helps in trajectories (closing 26 to 36% of the gap, CIs excluding zero); its sign reveals whether a benchmark's surface variation is adversarial or incidental. An LLM reranker recovers 5 to 63% of the gap in mathematics and 43 to 76% in trajectories; direction replicates across three judges (all 21 cells positive), but effect sizes, tier profiles, and the outlier judge change with domain (paired differences excluding zero everywhere). Mathematics gains concentrate on well-known competitions (+19.8 points, CI [+6.7, +33.2], one of six cells), so part of the recovery is memorization. In a paired downstream experiment (210 queries, graders at 96 to 99% agreement), oracle retrieval was indistinguishable from adversarially bad retrieval (McNemar p = 0.678); the solver's 69.5% zero-shot accuracy is largely a truncation proxy (97 to 100% on finished answers), leaving no headroom.
Aug 31, 2026

Learning from What You Retrieve: Online RL Fine-Tuning for Semantic Retrieval

In large-scale e-commerce retrieval, dual-encoder retrievers are op- timized for contrastive similarity, whereas downstream rerankers capture finer-grained relevance preferences; this objective mis- match limits end-to-end retrieval quality. Reinforcement Learning offers a way to use reward-model feedback for retriever adaptation, but we observe that standard policy-gradient updates can degrade embedding geometry, especially when the document index must remain frozen due to industrial constraints. To address this, we propose PAO (Positive-Advantage-Only), a selective RL optimization method. Our analysis reveals that in- discriminate penalization of negative samples (pushing away) in a frozen high-dimensional space disrupts pre-trained semantic man- ifolds. PAO selectively applies gradient updates only to retrieved items with positive advantages, effectively pulling query embed- dings toward high-reward regions while preserving global topo- logical stability. Experiments on both a massive industrial dataset and public benchmarks demonstrate that PAO significantly outper- forms standard RL and distillation baselines.
Aug 31, 2026

E-SENS: Exclusion-Sensitive Penalization for Negative-Constraint Retrieval

Retrieval-augmented language models can fail to respect negative constraints when the retriever supplies evidence about concepts the user explicitly excluded. Beyond explicit negation, queries may ask for answers that include one concept while excluding another, or for entities that belong to a category but differ from a closely related instance. Because the excluded concept still appears in the query text, dense retrievers may assign high similarity to documents about that concept even when the user asks to avoid it. We introduce E-SENS, a training-free reranking method for negation-sensitive retrieval. E-SENS extracts a compact trap query for the excluded side and subtracts trap-query similarity from the original-query retrieval score. On ExcluIR, E-SENS shows a clear recall-violation trade-off across four embedding models and reduces trap retrieval at recall-preserving settings.
Aug 30, 2026

REIGN: Refurbished Embeddings with Integrated Guidance Networks for Efficient Context-Length Scaling

Dense retrieval over long documents is expensive. Token-level encoders scale quadratically in sequence length, and most long-context embedding models reach 32K tokens only through architectural workarounds or by stretching billion-parameter LLMs. We propose REIGN (Refurbished Embeddings with Integrated Guidance Networks), a contrastively trained bi-encoder that operates on sequences of contextualised chunk embeddings from a frozen Guidance Network (GN) rather than on raw tokens. REIGN targets multi-chunk inputs, primarily for document-to-document retrieval; single-chunk inputs stay with the GN. Decoupling token-level processing from document-level reasoning, and caching the GN embeddings to disk, cuts per-document training cost by roughly four orders of magnitude relative to chunked Transformer fine-tuning. We also release a synthetic long-document retrieval benchmark for contrastive training and evaluation at long context lengths. Across an in-distribution Wikipedia benchmark, the LoCo out-of-distribution suite, and a real-world patent retrieval case study, REIGN matches dense long-context retrievers at smaller parameter budgets in each regime. A paired significance test puts it on par with models 1.6-4.3x larger on the patent task, and it stays within 0.65 nDCG@10 of a 20x-larger model on LoCo.
Aug 29, 2026

Efficient GPU Retrieval for Semantic Search

Semantic Search on LinkedIn must retrieve relevant profiles from a corpus of hundreds of millions in response to natural-language queries such as "a fintech founder in Berlin who worked in payments." The deployed relevance policy is bottleneck-oriented: every active non-negotiable facet must be satisfied, and a pre-existing LLM Graded Relevance (GR) judge operationalizes this through a fixed min/median aggregation over facet grades. Cosine similarity instead averages evidence, letting a strong match on one facet mask failure on another, capping the recall of the first-stage (L0) retriever. We present a policy-aligned retrieval framework: embeddings are partitioned into eight category-supervised segments whose scores follow the same min/median rule at serving time; for multi-vector retrieval, this segment score is computed independently per tagged document slot and maximized across slots. A lightweight single-slot Stage-1 scorer generates high-recall candidates, while scale-invariant relative-norm gating keeps category activation consistent across training, evaluation, and serving. On 21K held-out queries, this representation improves offline relevance over a matched-capacity baseline, with gains broadly distributed across facet combinations. We serve this framework with a two-stage GPU architecture: an FP8 coarse ranker scores the full corpus, increasing per-shard capacity by 71% and Stage-1 matmul throughput by 36%, then an FP16 stage exactly re-ranks an oversampled candidate set, recovering 99.6-99.8% of full-FP16 recall at over 500 QPS per shard replica. In a member-randomized A/B test, exploratory-query Precision@10 under the unchanged GR judge rises from 63.7% to 79.0% and navigational Precision@1 from 65.5% to 74.7%, with a blinded human evaluation independently confirming the Precision@10 gain.
Aug 26, 2026

Pointing the Way, Hiding the Destination: Practical Private Dense Retrieval at Scale

Hosted retrieval-augmented generation (RAG) and semantic search allow users to query valuable provider-held corpora, raising two competing demands: to hide each query and chosen result, yet reveal only the documents that the user is authorized to receive. Existing cryptographic approaches either make this costly by processing the entire corpus for every query, or sacrifice quality for efficiency by scanning a few clusters. We repurpose learned deep hashing as a private filter: a randomized binary code points the provider to a short candidate list, while encrypted reranking and oblivious key transfer protect the precise query and final selection. This shortlist short-circuits full-corpus cryptographic search without sacrificing retrieval quality: with 200-500 candidates, it closely matches full-corpus retrieval across five zero-shot corpora spanning 25K to 5.4M documents. On the full 2.68M-passage NQ corpus over a 10-Gbps link, our protocol only adds 0.38 seconds, or 6.5%, to a 128-token Qwen3-32B RAG pipeline. The released code satisfies directional metric differential privacy (DP) and substantially reduces embedding-inversion and property-inference leakage, demonstrating that a carefully learned shortlist can make private dense retrieval both accurate and practical.
Aug 25, 2026

Keeping the Index Open: The Recommendation-Side Cost of Shared Search and Recommendation

A shared search-and-recommendation index must score new items from features alone because search has no exploration slot. In a public log covering both surfaces over one catalog, 38.6%38.6\% of held-out query-search impressions show an item never previously shown or visited. For user-cold engagements, the feature-based tower serves this demand without measurable loss against 9999 sampled negatives (0.95950.9595 Recall@20 versus 0.95100.9510 warm). A lexical baseline reaches similar parity, while a full-catalog check remains statistically undecided. Dual-encoder retrieval therefore keeps the index open to new items, unlike an ID-softmax recommender that requires retraining. We price this openness on recommendation against six sequential baselines, each retrained and tuned through five rounds on corrected targets. A float32 timestamp bug had reordered leave-one-out targets for 19.7%19.7\% of users. On MovieLens-1M, warm accuracy trails the strongest retrained baseline by 5.2%5.2\% Recall@20 and 11.4%11.4\% NDCG@20. On MIND, the gap narrows to 0.80.8--3.6%3.6\% relative to the five strongest baselines, though the model ranks sixth of seven. Under strict zero-leakage cold-start evaluation, the content tower achieves 0.172±0.0060.172 \pm 0.006 Recall@20, 1.4×1.4\times the strongest retrained dedicated method (0.124±0.0070.124 \pm 0.007) and 3×3\times a training-free floor, without cold-specific training. Exact full-softmax training raises Recall@20 by 54%54\% on MIND-small and 6.9%6.9\% on MovieLens-1M over sampled InfoNCE, but recomputes the full catalog each step and exhausts accelerator memory at 240240K items. Approximate nearest-neighbor search explains none of the remaining gap, serving cost does not regress against ID-softmax retrieval, and a history-window sweep explains half the post-recipe remainder. Exact-quality training at catalog scale remains the open problem.
Aug 24, 2026

Hypergraph Embedding Indexing for Efficient Dense Vector Retrieval

Dense vector retrieval has become the foundation of modern semantic search, yet existing approximate nearest neighbor (ANN) indexes treat an embedding as an indivisible point in a high-dimensional space. In this work, we propose the Hypergraph Embedding Index (HEI), a framework that instead organizes documents according to combinations of highly activated latent embedding dimensions. This formulation enables inverted-index style candidate generation while preserving the semantic ranking capabilities of dense embeddings. We further demonstrate that constructing multiple complementary hypergraphs substantially improves retrieval coverage without the combinatorial growth associated with increasing the dimensionality of a single hypergraph. Finally, we establish that the statistical properties of embedding activations strongly influence coordinate-inverted indexing efficiency, introducing activation diversity as a diagnostic metric governing embedding indexability in coordinate-inverted frameworks.
Aug 21, 2026

Retrieval Needs Multivectors: An Exponential Separation

Recent works have highlighted the expressive limitations of embedding based retrieval models through both theoretical analyses and challenging benchmarks such as LIMIT. While multi-vector embeddings consistently outperform single-vector embeddings, the precise representational gap between them remains poorly understood. In this work, following Jayaram's work, we provide the first explicit family of query and document sets, together with their relevance matrices, for which single-vector embeddings that rank all relevant documents above irrelevant ones require exponential size, whereas polynomial-size multi-vector embeddings suffice. Our result establishes an exponential separation between the expressive power of single-vector and multi-vector embeddings for the task of ranking of documents as opposed to approximating numerical scores as in the work of Jayaram. Motivated by our theoretical construction, we introduce ANDOR, a new retrieval benchmark that naturally instantiates these hard examples. We show that state-of-the-art single-vector embedding models perform poorly on ANDOR in the zero-shot setting and exhibit only marginal improvements after fine-tuning, highlighting the inherent difficulty of the benchmark compared to prior work. In contrast, multi-vector models consistently outperform their single-vector counterparts and improve substantially with fine-tuning, closely aligning with our theoretical predictions.
Aug 20, 2026

Natural Language Code Retrieval for 1C:Enterprise: An Open Benchmark and Efficient Bi-Encoder

Natural language code retrieval is a rapidly evolving task in computer science. However, the 1C:Enterprise ecosystem combines Russian syntax with highly domain-specific terminology, for which open datasets and specialized models have been virtually non-existent. We present a comprehensive pipeline for 1C code retrieval: an open benchmark of 3,413 real-world, PII-scrubbed query-code pairs, a reproducible evaluation harness, and a specialized bi-encoder. To overcome scarce labeled data, we fine-tune on 784,057 synthetic triplets generated by google/gemma-4-26B-A4B-it from public code repositories, using Matryoshka Representation Learning (MRL) and a privacy-aware tokenizer. Because the benchmark subsets differ in size, we report balanced-subset macro, query-weighted micro, and forum-only results. Our model reaches 0.5992 balanced macro nDCG@10, 0.5044 micro, and 0.4617 on forum, versus 0.4932 macro for the baseline architecture and 0.5404 for google/embeddinggemma-300m. Removing every benchmark example flagged by the conservative exact/13-gram overlap audit leaves 0.6011 balanced macro (0.5010 micro), indicating that detected train-benchmark overlap does not explain the headline result. MRL truncation to 256 dimensions preserves 99.9% of retrieval quality while reducing dense-index storage and exact similarity arithmetic by a factor of three.
Aug 19, 2026

GreekBarRetrieval: A Benchmark for Greek Statutory Retrieval

Statutory retrieval is necessary for citation-grounded legal question answering, but remains underexplored for Greek. We introduce GreekBarRetrieval, a public retrieval benchmark derived from, and complementing GreekBarBench, which did not include retrieval. The new benchmark comprises 283 bar-exam questions, each accompanied by the facts of the case it refers to, and 6,308 candidate statutory articles to retrieve from. Questions and facts are stated in everyday language, but need to be mapped to the formal terminology of statutes and their abstract legal concepts. A further complication is that not all of the case facts are relevant to each question of a case. Experimenting with three BM25 variants and nine dense retrievers, we find that vanilla dense retrieval far outperforms vanilla sparse retrieval in Recall@100. However, LLM-based query reformulation helps BM25 close that gap, while also improving dense retrieval. With a ten-round ReAct-like LLM reformulation loop that we introduce, BM25 improves further in Recall@100 and obtains the best nDCG and MAP scores of all tested retrievers. Query reformulation also outperforms pseudo-relevance feedback, sparse-dense fusion, and English translation.
Aug 18, 2026

DEPT: Document Embedding Preservation Tuning for Unified Query Expansion and Retrieval

Large language models (LLMs) can both expand underspecified queries and encode text as dense representations, suggesting a unified model for query expansion and retrieval. Existing systems usually rely on prompted expansions, independently trained modules, or staged optimization, leaving generated expansions only indirectly aligned with the retrieval loss that judges them. We train a single decoder-only LLM end to end, where the same model generates the expansion and encodes both the expanded query and candidate documents. This unified setting creates a moving-target problem: retrieval supervision should improve query-side expansion, but the same update also shifts the document embeddings that serve as retrieval targets. We introduce Document Embedding Preservation Tuning (DEPT), which keeps tuned document embeddings close to cached initial embeddings while allowing retrieval gradients to pass through straight-through decoding into the generator. DEPT converts joint query--document movement into query-side adaptation against approximately stable, whitened document embeddings that support index reuse and online hard-negative mining. Experiments with Qwen3-4B-Instruct-2507 and LLaMA-3.2-3B-Instruct on five datasets in BEIR benchmark show that DEPT improves average retrieval quality over training-free, independently trained, and staged unified baselines, while ablations isolate the effects of preservation, whitening, end-to-end expansion training, and online negatives. Code is available at https://github.com/ILSparkle/DEPT.
Aug 17, 2026

The Commercial Tax: Rent-vs-Own Blind Spots in Multi-Hop Retrieval Benchmarks

Enterprises connect language models to their own data through retrieval. The benchmarks that rank multi-hop retrieval systems leave out two facts a buyer needs before a published number can be used: whether the retrieval backbone may be deployed commercially, and what it costs to build. On licensing: the field's dense-retrieval anchor, NV-Embed-v2, is licensed cc-by-nc-4.0. Of the four leading MuSiQue systems we audit (HippoRAG-2, PropRAG, SAG, KET-RAG), three depend on it for their best numbers and none says so. On performance: we measure thirteen embedders from eight makers on one identical MuSiQue harness with bootstrap confidence intervals throughout. Until mid-2026 there was a real commercial tax: the best commercially-licensed embedder trailed the anchor by 2.31 Recall@5 points (95% CI [0.91, 3.71], p=0.001). NVIDIA's Nemotron-3-Embed-8B, released 2026-07-16, has closed it: +0.24 at Recall@5 (95% CI [-0.94, +1.43], p=0.69), -0.58 at Recall@10 (p=0.28). It matches the anchor, does not beat it, and is the only entrant that is commercially licensed, free to self-host, and indistinguishable from the anchor; every other entrant meeting the first two conditions sits 5.2 to 14.6 points below. The durable finding is the paid-versus-free divide: API embedders charge per token on every re-index, self-hosted ones charge nothing. On cost: three of five audited systems (adding Microsoft's GraphRAG) do not disclose indexing cost, and the only published GraphRAG dollar figures span 11x inside one third-party paper (USD 2.30 vs USD 24.94 to index a 5.64 MB corpus once); extrapolated to 1 TB that undisclosed choice separates roughly USD 428K from $4.6M. Our cost model keeps one-time embedding apart from recurring answering: at 1 TB, embedding sits 7.5x-900x below graph construction, and a year of answering at 10,000 queries/day sits 350x or more below it.
Aug 16, 2026

Navigation-Informed Embeddings: Dense-Retriever Adaptation from Agent Search Traces

Agentic retrieval workflows produce query, retrieval, and stopping traces as a byproduct of answering questions. We study how these traces can adapt a deployed dense retriever to changing workflow distributions without new relevance labels, synthetic queries, or LLM judgments. We introduce Navigation-Informed Embeddings (NIE), a family of trace-derived objectives. NIE-Stop turns the stopping document into a soft positive; NIE-Path additionally uses preceding path documents as hard comparisons and imposes ordinal constraints with geometric decay. A BGE encoder adapted from retained source trajectories improves support Recall@20 on an independent target benchmark from 72.2 to 78.0 overall. NIE-Stop reaches 76.9 overall and 52.3 on long paths; NIE-Path raises long-path performance to 55.4, compared with 46.7 for the unadapted encoder. A shuffled-order control under the full path objective loses 3.2 points. Without public-benchmark training, the same adapter also improves nDCG@10 by 1.9 points on standard BEIR HotpotQA. NIE therefore provides a lightweight adaptation channel for settings where trajectories are already retained, with zero incremental labeling cost.
Aug 16, 2026

Query Expansion Should Be Coordinated: Dense Expands, Sparse Anchors

Retrieval-augmented generation (RAG) systems rely on retrieval modules to ground large language model (LLM) outputs. LLM-based query expansion enriches retrieval with document-like passages, but evaluations of hybrid retrieval often fuse fixed top-L prefixes of dense and sparse rankings. Because L controls cross-channel contributions and ranking access, it can alter measured expansion gains. We therefore evaluate complete-list effectiveness and record per-channel replay stopping depths required to certify the ordered top-K. This changes the design: because both rankings determine the fused result, their query constructions should be coordinated rather than designed independently. We present DESA (Dense Expansion and Sparse Anchoring), which shares generated references across channels but specializes their integration. Orthogonal residual expansion adds new semantic directions to the dense query, whereas score-product anchoring reorders the original sparse support without admitting expansion-only matches. The same references thus play complementary roles: Dense expands; Sparse anchors. Across seven BEIR datasets, DESA improves nDCG@10 and Recall@20 over the unexpanded query by 3.82% and 2.38%, while reducing dense and sparse replay stopping depths by 36.90% and 36.56%.
Aug 14, 2026

Where Does Retrieval Fail? Evaluating RAG Architectures for Agricultural Advisory

Retrieval quality in RAG systems is commonly reported as a single aggregate score, which can hide large differences across query types and language conditions. We study this problem in Bengali agricultural advisory, where farmer queries are often colloquial while official advisory documents use formal scientific terminology. We construct a test collection of 1,000 queries and 2,882 knowledge nodes extracted from 284 official Bangladeshi agricultural publications, and use it to evaluate five retrieval architectures and six embedding models under three controlled language conditions. The results show that no single retrieval method is consistently best. For native Bengali queries, BM25 is the strongest single retriever (R@10 = 0.506) while Hybrid RRF reaches the highest overall R@10 of 0.539. However, dense retrieval performance varies sharply by query type: R@10 is 0.093 on colloquial farmer queries and 0.970 on formal safety queries. Across language conditions, BM25 R@10 drops from 0.506 on Bengali queries to 0.004 when English queries are matched against the Bengali corpus, while dense retrieval falls only from 0.464 to 0.425. We also find that embedding task configuration and passage length can each change reported R@10 by a factor of seven, independent of architecture. These results show why low-resource RAG evaluation should report performance by language condition and query type rather than relying on aggregate scores alone. The dataset and evaluation scripts are available at https://huggingface.co/datasets/RaiyanKhaan/AgriTrust-RAG.
Aug 14, 2026

Retrieval Grounding Latent Reasoning for Dense Retrieval

Reasoning-intensive retrieval requires text representations to capture not only semantic similarity, but also the reasoning needed to determine relevance under a given retrieval instruction. Existing reasoning-enhanced embedding models improve retrieval by incorporating reasoning information into dense representations, yet their supervision is typically dominated by the final retrieval objective. As a result, latent reasoning trajectories may learn shortcut reasoning patterns that preserve retrieval performance without producing meaningful incremental retrieval gains. We propose Retrieval Grounding Latent Reasoning (RGLT), a latent reasoning framework for dense retrieval that explicitly connects intermediate latent transitions with retrieval improvements. RGLT performs non-autoregressive reasoning in hidden space through an instruction-conditioned latent reasoning trajectory constructed from silent tokens. It combines process-supervised explicit-to-implicit distillation with retrieval-grounded supervision, using stage-wise CoT reconstruction to shape intermediate latent states and retrieval-effect credit to optimize incremental retrieval gains across the latent reasoning trajectories. Experiments on reasoning-intensive retrieval benchmarks show that RGLT consistently outperforms strong baselines while preserving efficient embedding inference.
Aug 11, 2026

DistilVDR: A Compact End-to-End Visual Document Retriever via Dual-Student Distillation

Visual document retrieval (VDR) is dominated by multi-billion-parameter models that are slow to index at full corpus scale and expensive to serve. Prior compression routes either train a smaller multi-vector encoder from scratch or distil only the query side; neither yields a compact single-vector retriever end-to-end. We present DistilVDR, a 524M end-to-end VDR system distilled bilaterally from a single 8B vision-language teacher under a pointwise cosine alignment loss. All supervision comes from the frozen teacher's embedding space, which was itself trained with relevance supervision, so the student objective needs no relevance labels, negative sampling, or contrastive term. We match VDR's text-query and image-document input asymmetry with an asymmetric encoder-only student that concentrates visual capacity on the document side and keeps the query side at 70M parameters. We release two variants that share the same encoders and training and differ only in the document encoder's visual-tile budget: DistilVDR-HiRes attains 61.74 average NDCG@5 on ViDoRe v1+v2+v3 (86.9% of the 8B teacher) and leads every reproduced sub-1B baseline on the high-resolution-sensitive v3 benchmark, while DistilVDR-Fast attains 59.98 at a 3 times smaller visual-token budget. Both variants store one million documents in a 15.6 times smaller index than the strongest sub-1B multi-vector baseline and index the corpus an order of magnitude faster. The code is available at https://github.com/Ryenhails/NanoVDR.
Aug 5, 2026

Teaching Nemotron Greek: Mining a Corpus, Adapting Retrieval, and Grounding Generation for Modern Greek across Specialist Domains

Modern Greek is absent from NVIDIA's Nemotron retrieval models and from major multilingual retrieval benchmarks, despite being important for retrieval-augmented generation (RAG) in legal, energy, financial, and medical applications. We present an end-to-end adaptation of the Nemotron retrieval stack for Modern Greek, including corpus mining, synthetic supervision, retrieval model training, reranker adaptation, reader fine-tuning, and a new benchmark called HERA. Our study shows that a parameter-free BM25 baseline outperforms several off-the-shelf multilingual dense retrieval models on specialist Greek corpora. After fine-tuning on 65,773 Greek retrieval pairs, a Nemotron 1B embedder improves nDCG@10 from 0.362 to 0.835 and substantially outperforms its unadapted counterpart. The learned language competence transfers to general-domain Greek, although the advantage over BM25 remains domain-dependent. We further adapt a cross-encoder reranker and demonstrate consistent improvements across specialist domains. Finally, we LoRA-tune a Nemotron 30B-A3B mixture-of-experts reader for grounded generation, increasing judged answer correctness from 29.4% to 66.9% while significantly improving faithfulness and citation quality. We also introduce HERA, the first large-scale Greek benchmark for retrieval-augmented generation, and release our adapted models and benchmark to support future research on Greek-language RAG systems.
Jul 31, 2026

Bridging the Question-Answer Gap in Retrieval-Augmented Generation: Hypothetical Prompt Embeddings

Retrieval-Augmented Generation (RAG) systems synergize retrieval mechanisms with generative language models to enhance the accuracy and relevance of responses. However, bridging the style gap between user queries and relevant information in document text remains a persistent challenge in retrieval-augmented systems, often addressed by runtime solutions (e.g., Hypothetical Document Embeddings (HyDE)) that attempt to improve alignment but introduce extra computational overhead at query time. To address these challenges, we propose Hypothetical Prompt Embeddings (HyPE), a framework that shifts the generation of hypothetical content from query time to the indexing phase. By precomputing multiple hypothetical prompts for each data chunk and embedding the chunk in place of the prompt, HyPE transforms retrieval into a question-question matching task, bypassing the need for runtime synthetic answer generation. This approach does not introduce latency but also strengthens the alignment between queries and relevant context. Our experimental results on six common datasets show that HyPE can improve retrieval context precision by up to 42 percentage points and claim recall by up to 45 percentage points, compared to standard approaches, while remaining compatible with re-ranking, multi-vector retrieval, query decomposition, and other RAG advancements
Jul 29, 2026

Models for minimalist RAG: B1ade 335M Embedding and 1B Parameter Small Language Models

Language and embedding models used in RAG systems are conventionally assumed to require large-scale pretraining and explicit grounding supervision. We present B1ade, an efficient RAG architecture comprising two purpose-built components: a compact embedding model and a purpose-built SLM. B1ade-embed, a 335M parameter retrieval model constructed via parameter-free fusion of five pretrained encoders achieves top MTEB scores among sub-500M models with zero additional training, and B1ade-1B, an SLM trained on low-cost GPUs using Group Relative Policy Optimization (GRPO) on 723M tokens (2.2M examples) of curated context-question pairs with rewards that optimize only answer similarity. Our central finding is emergent attribution: despite receiving no explicit supervision for source citation, B1ade-1B cites retrieved passages in 42.4% of responses, exceeding the attribution rate of its training distribution by 5.5 percentage points. This demonstrates that grounding behavior can emerge as an accuracy-maximizing strategy under RL training, without explicit reward engineering. On standard QA benchmarks, B1ade-1B achieves 81.82% on PopQA, 65.8% on PubMedQA, and 51.09% on FEVER. In end-to-end RAG evaluation, B1ade-1B achieves an average score of 0.654 across correctness, completeness, coherence, and faithfulness, a 10.8% improvement over the SFT, while closing the gap with models 1.5x its size. These results show that strategic model composition and reward design suffice for resource-efficient RAG, without large-scale pretraining.