New Information Retrieval (cs.IR) papers, Wednesday 21 October 2026
21 papers announced on arXiv. Get them ranked for you, free →
MOSS: A Production Framework for Monitoring and Diagnosing Semantic Failures in Search and Recommendation Systems
AI summaryMOSS integrates LLM relevance scoring into a production monitoring loop, sampling data, prompting strategically, and alerting with diagnostic root‑cause categorization. In a large retailer, it detected two defects and quantified a 76.4% alert‑rate drop, outperforming non‑LLM baselines on recommendation and search benchmarks.
Expert-Guided Rubrics as Dense Reward Signals for Reinforcement Learning in Non-Verifiable Domains
AI summaryUses expert‑rated rubrics to turn sparse, graded SME feedback into dense rewards for RL agents in non‑verifiable retrieval tasks; the hybrid signal boosts Recall@3 by 6% and MRR@3 by 6.7% over sparse‑reward baselines.
GRaSP: Latency-Aware Optimization for Geo-Distributed Retrieval-Augmented Generation
AI summaryGRaSP adds a latency‑aware re‑scoring layer to RAG that penalizes distant passages and optional costs, solving a matroid‑constrained optimization in O(N log N). It cuts P90 fetch latency by 41% and keeps 78 % of relevance recall with negligible overhead.
The Bottleneck Is Retrieval, Not Reading: Diagnosing Evidence-Based Generation with RAG for Long Context Question Answering
AI summaryThe paper proposes evidence‑based generation (EBG), requiring line‑level citations in retrieval‑augmented QA, and a joint Ans+Evid metric. It shows that similarity‑based retrieval fails to surface all needed chunks for multi‑fact queries and that chunking destroys document‑level context needed for identity‑bearing questions.
MultiCard: Multiple Attribute Cardinality Estimation via Deep Learning
AI summaryMultiCard is a deep learning estimator that unifies string LIKE predicates and numerical filters for database query optimization. It reduces PostgreSQL's mean Q-error by up to 16 times on mixed workloads while keeping inference latency below 5 ms.
Joint Constraint Aggregation for Qualifier-Sensitive Retrieval-Augmented QA
AI summaryJoint Constraint Aggregation (JCA) compares candidates against the full set of question-derived requirements using retrieved evidence, reducing bias toward frequently mentioned partial matches. Across three datasets and instruction-tuned models, JCA improves accuracy, especially when retrieval favors competing answers.
UIBK-Embedding: Effective Text Retrieval by Unifying Dense and Late-Interaction Retrieval
AI summaryThe authors propose UIBK-Embedding-v1, a 179M-parameter encoder that unifies dense and late-interaction retrieval into a single representation and index. It achieves state-of-the-art BEIR performance for its size class by combining token-level matching and dense cosine similarity in one pass.
Neighborhood-Aware Unlearning for Generative Recommendation
AI summaryThis paper identifies that unlearning in generative recommenders causes unintended utility degradation in the semantic neighbors of forgotten items. To fix this, the authors propose Neighborhood-Aware Unlearning, which explicitly controls probability redistribution and successfully restores local recommendation performance.
FLINT: Fixed-budget Late Interaction
AI summaryFLINT introduces a fixed-budget late interaction retrieval architecture that applies symmetrical attention-based pooling to both queries and documents. It reduces index size by up to 9x and accelerates retrieval compared to token-level models while maintaining competitive ranking effectiveness.
SweepANN: A GPU-Accelerated Large-K Approximate Vector Search for Drug Discovery
AI summarySweepANN is a GPU-accelerated approximate nearest neighbor search algorithm optimized for large-K retrieval in drug discovery. It achieves 99.6% mean recall on a 500-million molecule dataset in about 61 seconds, outperforming traditional IVF and graph indexes at high K.
Conversational AI for Perceived Cognitive Decline Screening in Older Adults
Early detection of cognitive decline in older adults remains challenging due to the resource demands of standard neuropsychological assessments. Chatbot-based Intelligent Cognitive Assistants (ICAs) provide a scalable alternative by eliciting open-ended responses rich in…
Baseline and Incremental: User, Item, and Request Debiasing in Instagram Feed Ranking
AI summaryThe paper proposes keeping raw engagement predictions and adding a separate incremental term in the value model for Instagram feed ranking. A four‑arm ablation shows the two‑term design outperforms single‑term versions in most metrics, and the operator is now widely used across Meta apps.
Factual Access States in Diffusion Language Models: From Mechanism to Knowledge Editing and Retrieval
Factual recall is a lookup: some representation inside the network serves as the key, and what that key is made of decides which queries can retrieve the fact, what an edit must cover, and what a retriever can read. We call it the factual access state. In autoregressive (AR)…
GARRET: Generative Feedback for Region Retrieval in Visual Documents
AI summaryGARRET trains a region-level visual retriever using only (page, question, answer) triples via differentiable generative feedback, avoiding expensive region labels. It achieves 0.4126 macro‑nDCG@10, 2.94 points above baselines, while using 47.8% fewer page vectors.
Robustness Analysis of Pareto-based Joint Evaluation of Fairness and Relevance in Recommender Systems
AI summaryThis paper introduces Distance to Pareto Frontier (DPFR), a modular evaluation approach that jointly measures recommendation relevance and fairness using Pareto-optimal trade-offs. Experiments show that DPFR provides a more robust and reliable evaluation than existing inconsistent proxy measures.
SPAR: Stepwise Personalized Autoregressive Recommendation with Prefix-Balanced Semantic IDs
AI summaryThe paper introduces SPAR, a generative recommendation framework that uses step-by-step personalized decoding and balanced semantic IDs to prevent branch-selection errors and popularity bias. Experiments on Youku production data show improved hit rates and user engagement metrics.
AutoSim: A Self-Improving User Simulator for Conversational Search Evaluation
AI summaryAutoSim is a conversational‑search user simulator that self‑tunes its weights from real user signals, achieving realistic behavior: it ranks retrieval systems in strong agreement with NIST (τ = 0.66, 0.91, 0.91) and is indistinguishable from real users, reducing bias in system evaluation.
DeclineCraft: A RAG Abstention Benchmark for Jointly Measuring Combined Retrieval and Parametric Knowledge
AI summaryThe authors introduce DeclineCraft, a benchmark for evaluating when retrieval-augmented generation systems should abstain from answering unanswerable questions. They find that advanced retrieval strategies and reasoning modes paradoxically reduce abstention rates by prompting models to fabricate answers or deny premises rather than acknowledge missing knowledge.
Multimodal Reranking Is a System One Decision
AI summaryA general VLM can rerank multimodal retrieval candidates by treating reranking as a single “choice” decision: ask the model to pick the best among the top‑16 candidates. This zero‑shot approach outperforms specialized rerankers on image‑retrieval benchmarks and rivals larger supervised models with similar latency.
Evolving Semantic Indexes for Generative Retrieval
AI summaryThe paper uses LLM-driven evolutionary search to mutate generative retrieval components, guided by a fast 1M‑parameter proxy decoder. Across 21 SID programs on NQ320K, evolution boosts Recall@1 from 55.8% to 59.3% and Recall@100 from 85.8% to 90.6%.
ECHO: Efficient Cascade with Homogeneous Operators for Distillation-Guided 100K-Sequence Modeling
AI summaryECHO efficiently models 100K-long user sequences by selecting a subset of events via a distilled selector and retrieving target‑relevant records with a fine‑grained retriever, cutting online FLOPs by 160× and boosting GAUC by up to 1.5 % over SOTA.