arXane › Computer Science › Computer Science
Wednesday, 4 November 2026: 21 new papers, 5 cross-lists and 9 replacements in Information Retrieval (cs.IR), in arXiv's order, with full abstracts. The same papers as arXiv's own listing, set as a newspaper; every title links to arXiv.
Next point-of-interest (POI) recommendation supports personalized place discovery on location-based Web platforms. Mobility routines can recur across users while being realized at different places and adapted to individual preferences. We present DualClusterNet, a causal sequential recommender that composes two population-shared prototype dictionaries. At each observed prefix, soft routers retrieve pattern and spatial mixtures; a factorized multiplicative interaction couples them, feature-wise affine modulation adapts the shared representation, and a separate user-conditioned branch with an additive--projective gate retains individual variation. The prediction head combines this representation with a learned first-order transition term and ranks the complete training POI vocabulary. We evaluate on NYC and TKY using a training-only vocabulary and an original-final-target protocol that neither bridges unknown POIs nor substitutes earlier targets. The available single-seed results show competitive full-candidate ranking, while controlled analyses support the contribution of the compositional and personalized branches. Code is available at https://anonymous.4open.science/r/DualClusterNet-F4EE/.
Long-term memory enables large language model (LLM) agents to provide continuous personalized assistance by retaining user information across interactions. Existing systems support this process through fact extraction, organization, updating, and retrieval, but largely maintain memories as independent facts. This fact-centric paradigm may overlook how new facts update, constrain, or conflict with earlier ones, while leaving their applicability, role, and priority for the current request implicit. In this paper, we introduce EvoMemState, a state-centric framework that augments persistent facts with two complementary states: Relational States that capture how facts remain valid and relate to one another as memory evolves, and a Reasoning State that determines how these facts should be applied to the current request. Specifically, EvoMemState consists of three key components: (i) Ontology-Guided Memory State Induction extracts Persistent Facts and constructs Relational States that maintain their validity, scope, and evolving relations. (ii) RL-Based State Evolution internalizes memory-use experience into a policy through reinforcement learning to generate a request-specific Reasoning State. (iii) State-Aware Reasoning then combines the relevant facts with their Relational States and the generated Reasoning State to guide the personal agent. Extensive experiments on long-term dialogue and memory-use benchmarks demonstrate consistent improvements over existing memory systems in retrieval completeness and memory-grounded response accuracy. Our code and implementation details are available at https://github.com/XMUDeepLIT/EvoMemState.git.
Natural Language Processing (NLP) has been widely applied in financial domains, enabling automated analysis and retrieval of stock market information. However, traditional NLP-based systems often struggle with maintaining contextual accuracy and generating relevant responses in dynamic financial environments. To address these challenges, Retrieval-Augmented Generation (RAG) models have emerged as a powerful solution, combining retrieval-based methods with generative capabilities to enhance response quality. In this study, we present a Retrieval-Augmented Generation (RAG) system designed to generate responses related to stock market information. The system integrates retrieval-based and generative approaches to enhance the accuracy and relevance of financial text generation. To evaluate its effectiveness, we employ the ROUGE metric, specifically ROUGE-1, ROUGE-2, and ROUGE-L, to measure lexical and structural similarity between generated and reference texts. Experimental results on a dataset of 100 stock market-related queries demonstrate high ROUGE scores, indicating strong alignment with reference documents. Additionally, we assess latency to evaluate the system responsiveness, achieving an average response time of 1.16 seconds. These findings confirm that our RAG model is capable of providing accurate and timely stock market insights, making it a promising approach for financial information retrieval.
Approximate Nearest Neighbors (ANN) search over neural embeddings powers modern retrieval systems such as Retrieval-Augmented Generation. Embeddings come in three representations (dense, sparse and multi-vector), and efficient search combines an index structure, a similarity function and a quantization scheme. Existing libraries fix one combination, hard-code each one separately, or factor only one side of the problem, and none covers all three representations. We present Vectorium and kANNolo, two open-source Rust libraries that split this design space into a data-management layer and an indexing layer. Vectorium abstracts the vector, the distance function, the quantizer and the dataset independently, and provides 11 encoders, from 32-bit floats to sub-bit codes. On top of it, kANNolo implements Hierarchical Navigable Small-World (HNSW) graph and Inverted Vector File (IVF) indexes that are generic over representation, encoder and metric, and a two-stage index that composes any first-stage index with any rescoring dataset. Since every combination is specialized at compile time, this modularity adds no run-time overhead, and the resulting indexes reach state-of-the-art performance: our Product Quantization matches Faiss in accuracy at competitive speed, our RaBitQ and HNSW match or outperform their reference implementations, and our IVF outperforms Faiss. Against six dense and five sparse systems on seven public collections, our indexes give the best memory-throughput trade-off on dense data in 9 of 12 collection-accuracy pairs and perform well on out-of-distribution queries. On sparse data, the systems built on Vectorium lead on all three collections, with Seismic and SeismicWave the fastest and kANNolo's indexes the most compact.
Generative search engines increasingly shape how users discover products, services, and commercial entities by generating answers from retrieved web content. This creates an opportunity to influence which brands appear in these answers by modifying the content available to the retrieval system. In this work, we study an adversarial generative engine optimization (GEO) setting in which a predefined brand is added to an article through a single contextually integrated mention. The challenge is to make the mention fit the surrounding article without making the content overtly promotional, while keeping it relevant enough to be retrieved and to influence the generated answer. To study this, we introduce Implicit Brand Injection (IBI), a multi-agent framework that identifies insertion locations, rewrites the selected sentence, verifies the edit, and selects the final candidate. To evaluate IBI at scale, we construct IBIBench, which contains 1,000 web articles and 60 synthetic brands across 12 diverse domains, with both endemic and non-endemic brand placements. We compare IBI with Direct Prompting and Best-of-N with ranking under an equivalent budget, evaluating insertion quality, target-chunk retrieval, and downstream brand propagation. Compared with the baselines, IBI produces insertions that better preserve semantics and disrupt perplexity less, while improving Hit@k across all evaluated cutoffs, query regimes, and placement settings. IBI achieves up to 82.2% Hit@10 and 0.74 MRR@10, while Brand Mention Rate (BMR) and Conditional Propagation (CP) reach up to 62.7% and 75.9%, respectively. These results show that a single brand mention can increase both retrieval exposure and brand visibility in generated responses, highlighting the need for defenses against implicit content manipulation in generative search.
Traditional recommender systems rely on arbitrary integer IDs to identify items, which carry no semantic information and force models to learn representations from interaction data. While leveraging content information can partially mitigate cold-start, this paradigm inherently suffers from exploding embedding tables and scenario-specific architectures. Recent years have seen the emergence of semantic IDs, which encode semantic information directly into the item identity, reflecting content, hierarchy, and relationships. This framework addresses the above limitations by offering a principled way to encode large-scale semantic information into item identity, benefiting both generative and discriminative applications and unlocking more native unification across modalities, domains, and tasks. This tutorial traces the evolution from raw IDs to semantic IDs as a scalable means of information utilization, offering a comprehensive review of their development and applications across multi-modality, cold-start, and cross-domain recommendation. Hands-on exercises guide attendees through building and evaluating semantic IDs, training models, and exploring cross-domain transfer. We conclude by discussing open challenges, including dynamic semantic IDs and multi-stakeholder utility. All materials of the tutorial could be found at the official website https://semantic-id-tutorial.github.io.
Large language models (LLMs) increasingly shape recommendations across chatbots, e-commerce, and user-generated content (UGC) platforms, creating demand for Generative Engine Optimization (GEO) to improve item visibility in generated responses. Existing GEO studies are mainly content-centric, optimizing visibility under general-purpose queries. However, a shared item description usually serves users with different needs and decision goals, resulting in its visibility gains being dependent on the request. To investigate how changing intent affects visibility, we formulate IntentGEO to evaluate shared rewrites across need specificity and decision progression. We revisit existing GEO methods and find that their visibility gains vary across user intents. More critically, learning-based prompt optimization can underperform its handwritten seed on average. Tracing these updates reveals prompt drift from explicit recommendation instructions toward generic promotional tactics. To address this drift, we propose AMO (Anchored Meta-Optimization), which preserves recommendation instructions in a fixed core while learning appended strategies from citation feedback. Across four domains and ten intent states, AMO improves average citation-rank gain by 35% over the strongest prior learning-based baseline and leads on nine states. Ablations show that learned additions improve on the fixed core alone and that AMO outperforms full-prompt optimization initialized from that core. It also achieves the highest gains in an agentic recommendation pipeline and, after separate adaptation, on Doubao, a commercially deployed, closed-source online assistant. These findings highlight user intent as a key dimension of GEO evaluation and support anchored adaptation across recommendation needs and decision stages.
Larger generative retrieval models can improve retrieval quality, but their decoding cost makes low-latency serving difficult. When different users submit the same query, generating complete Semantic IDs (SIDs) for every request repeats computation, while their interaction histories call for different results. The challenge is to reuse large-model computation across requests while keeping retrieval personalized. We propose \textbf{ReuseGR}, which separates query-level candidate planning from request-level personalized completion. A 7B offline planner generates SID prefix sets shared across requests with the same query. A 0.5B online model uses each request's history to select prefixes and generate the remaining tokens. ReuseCut chooses a cache depth for each query based on historical coverage, balancing computation reuse and candidate coverage. To strengthen online completion, we transfer 7B suffix-scoring knowledge to 0.5B through path-level knowledge distillation (Path KD). BudgetCut then selects scored branches to improve final top-100 recall under a fixed branch budget. Its selection policy is initialized by on-policy distillation (OPD) and refined through group-relative policy optimization (GRPO). On an industrial search-advertising dataset with approximately 2M query--item records and 3.38M catalog items, ReuseGR improves query-level HR@100 by 3.83 percentage points and reduces mean online latency by 38.9\% relative to an all-online 0.5B baseline when prefixes are prepared. It achieves HR@100 comparable to an all-online 7B baseline with 57.7\% lower online latency. A 14-day user-level online A/B test against the 0.5B baseline shows relative gains of 0.93\% in click-through rate and 1.05\% in revenue per thousand requests.
Semantic retrieval models aim to match documents to a query's "meaning", yet it remains unclear whether they successfully encode temporal relations between events in their ranking mechanisms. We investigate this question by studying timeline-constrained document retrieval, where a query is an ordered sequence of natural-language event statements while a relevant document should match those events in a compatible temporal progression. We then test whether widely used dense retrievers and neural rerankers capture such constraints, and whether explicitly preserving temporal order improves ranking, using a controlled setting over a novel focused evaluation collection. Experiments show that state-of-the-art semantic retrievers and rerankers provide only limited improvements over lexical retrieval, suggesting that their semantic matching capabilities do not reliably capture temporal constraints. We then compare a timeline-aware method with both semantic matching and standard rerankers and find that it yields substantial improvements. This demonstrates that considering temporal order can provide a strong ranking signal in this setting, suggesting future research on improving temporal modeling in semantic retrievers.
Recommender systems are widely used in e-commerce and content platforms, where improvement requires repeated model design, training optimization, and experimental validation. Large language model (LLM) agents increasingly automate this development cycle. Sustained autonomous improvement requires agents to improve both recommendation models and their development workflows. However, task-performance-driven approaches select updates mainly by candidates' final scores, without a separate optimization signal for experimentation quality. We therefore present DualEvolve, a dual-level framework that co-evolves recommendation mechanisms and agent orchestration toward recursive self-improvement. Inner evolution revises model and training code using task feedback to improve recommendation performance. Outer evolution scores the inner-loop experimental process, updates harness and skill rules using these scores and execution traces, and adjusts orchestration choices based on next-round scores to improve the quality of subsequent experiments. Experiments on MovieLens-1M and three Amazon Reviews 2023 subsets cover four recommendation methods and 16 settings. DualEvolve achieves higher test means than Base, OpenEvolve, and DeepEvolve in all 64 setting-metric comparisons, with macro-averaged relative gains over DeepEvolve of 7.6% (HR@5) and 9.0% (NDCG@5).
Diffusion-based recommenders learn user preferences by denoising interaction histories, but the added noise can distort the original preference signals. Flow-based collaborative filtering (CF) instead learns a discrete path from a sampled source to the user's observed training interactions (the target history), enabling inference in a few steps. Existing flow-based CF methods sample sources using global item frequencies to reflect the sparsity and popularity of observed interactions. Each source is sampled independently of its target history, leaving the predictor without user-specific source cues for learning to reconstruct that history. We propose SourceFlow, which brings collaborative structure into the source distribution itself rather than adding further conditioning to the predictor. Like inference from observed histories, training uses collaborative evidence to predict missing interactions. Specifically, Neighborhood Coupling constructs displaced binary sources by replacing historical items with unobserved collaborative alternatives. Collaborative Consensus further refines replacement probabilities using evidence from other interactions of users who share the observed item. Stretched Restoration retains these replacements at late flow times during training, providing more opportunities to recover the original interactions. Experiments on five recommendation datasets show consistent improvements over competitive baselines, including a mean relative Recall@10 gain of 15.7% over the strongest flow-based baseline. Controlled experiments confirm that these gains depend on correctly pairing collaborative sources with their target histories, as reassigning the same source vectors across users reverses the gains. Code will be publicly released upon acceptance.
While conversational recommendation systems provide a more natural user interface, they face the challenge of progressively discovering a user's needs before they can recommend an item that satisfies them. To evaluate the ability of AI agents at this elicitation process, we introduce $\tau$-Rec-Books, a benchmark that holds each user request fixed while varying only when the user reveals their requirements: all at the start (volunteer), one at the start and the rest when asked (mixed), or some never (hidden). A hidden requirement can be found only by proposing books and inferring it from the user's accept or reject verdicts. We generate 450 tasks over a catalog of 20,000 books from Wikidata, develop a rule-based simulated user to answer the agent, and score recommendations against catalog metadata without an LLM judge. Across four agents and two search interfaces, success rates fall for every agent and interface, on average from 0.975 (volunteer) to 0.756 (mixed) and 0.283 (hidden). Mixed tasks fail mainly because agents do not ask, and when told to ask, every agent reaches at least 0.931. On hidden tasks, most failures come from agents recommending a book they never proposed to the user. When told to propose first, agents only match a simple baseline that proposes popular books, and stay well below one that uses rejection to rule out possible requirements. We release the catalog pipeline, task generator, simulated user, validators, reference agents, and all agent traces at https://github.com/nbharaths/tau-rec/tree/main/data/tracks/books-wikidata-v2.
Collaborative search enables users to address complex information needs jointly. Advances in large language models (LLMs) enable autonomous agents to support collaborative information seeking within shared communication spaces. However, little is known about how these agents shape search experiences, behaviors, and outcomes compared with conventional search tools. We conduct a user study with 80 students working in 40 pairs. Each pair completes four collaborative search tasks under four scenarios: QAAgent, which answers queries using an LLM's parametric knowledge; CoSearchAgent, which uses multi-user conversational context to reformulate queries, retrieve web information, and generate source-grounded answers, while asking clarifying questions when needed; an in-chat search plugin that returns result lists; and manual Google Search. After each task, participants complete several questionnaires assessing their search experience and independently write a report; after all tasks, they participate in individual interviews. Results show that LLM-based agents reduce perceived workload, support collaboration, and reshape communication and evidence-use practices. They also encourage more independent paraphrasing, albeit at the cost of report richness. To the best of our knowledge, this is the first systematic empirical comparison of LLM-based agents and conventional tools for collaborative search across search experience, user behavior, and collaborative outcomes.
Multi-hop retrieval-augmented generation (RAG) supports Web information seeking and fact verification by recovering complementary evidence across related pages within a limited retrieval budget. Scoring evidence independently against the original question can miss supporting units whose relevance emerges through intermediate facts. Iterative retrieval incorporates these facts into subsequent queries, while graph retrieval explicitly models relations among evidence units. Yet fixed-query graph retrievers update entity representations without adapting the query condition to accumulated relational context. We view multi-hop evidence recovery as the coupled evolution of evidence representations and query conditions, with accumulated context informing both what the retriever represents and how it searches for complementary evidence. QueryFlow realizes this perspective by recurrently integrating query-conditioned summaries of graph context into a shared query state, which controls boundary inputs and relation features for subsequent propagation. Across four representative multi-hop question answering (QA) datasets, QueryFlow achieves macro Recall@10 (R@10), Full Recall@10 (FR@10), and LLM-Acc gains of 4.98, 8.40, and 8.0 percentage points over representative baselines, respectively.
Proactive query recommendation aims to suggest relevant search queries to users on the homepage of local service apps. Unlike conventional recommendation over a fixed item set, this requires open-set generation from a vocabulary of millions, where outputs must simultaneously align with user preference, match fine-grained spatiotemporal context, cover long-tail intents, and provide diverse alternatives within a few candidate slots. Existing approaches address these axes in isolation: standard tokenization suppresses long-tail queries, coarse location encoding misses fine-grained geographic demand, generic preference optimization fails to align with production impression-click distributions, and standard beam search collapses to near-duplicate outputs. We present SPADE (Spatiotemporal Preference Alignment and Diversity Enhancement), an end-to-end generative framework that jointly optimizes the generation space, contextual representation, preference distribution, and output slate quality through four components: Objective-Aligned BPE (OA-BPE) aligns subword merging with business value signals to boost long-tail query generation probability; Spatio-Temporal Adaptive Representation (STAR) replaces coarse location tokens with continuous multi-scale encodings and soft gating for precise spatiotemporal grounding; MARS-DPO (Multi-dimensional Adaptive Reward Sampling DPO) combines cross-domain pre-training with preference alignment on production impression-click data via composite rewards and adaptive hard negative sampling; and Sigmoid Diversity Beam Search (SDBS) improves intent coverage without sacrificing recall. Offline experiments on large-scale production data show that SPADE outperforms the strongest generative baseline by +5.6% HR@5, -40.3% GeoC@5, and +1.9% Exec@5. A one-week online A/B test confirms production impact: +1.88% Module CTR, +0.177% Search Volume, and +0.105% Search DAU.
Industrial recommender systems allocate substantial computation to pointwise candidate scoring, yet separate reranking and mix-ranking stages still assemble the final exposure slate using hand-crafted rules and specialized low-capacity models. Mix-ranking combines product cards, short videos, livestreams, and advertising cards supplied by distinct upstream business channels. Although these rules and models encode valuable domain knowledge, they are inherently limited in leveraging additional data and computation. They constrain the combinatorial space of candidate slates and determine objective trade-offs through manual tuning. To enable scalable and effective slate generation, we reformulate reranking and mix-ranking as a unified learning problem whose performance can improve with computation. Specifically, we instantiate this formulation across the reranking and mix-ranking stages and propose OneTrans-GSR, a single decoder-only policy that extends our deployed OneTrans backbone to generate the exposure slate end-to-end. During training, a reward model evaluates complete slates and guides reinforcement learning toward slate-level and long-term objectives. During decoding, business constraints are explicitly enforced. Offline evaluation shows that slate quality consistently improves as reward-model and policy compute increase. Deployed in a large-scale industrial recommendation system, OneTrans-GSR mixes heterogeneous content types within a unified slate. In online A/B tests, it improves GMV/u by 2.63%, advertiser value by 2.06%, and repurchase rate by 2.03%, while reducing end-to-end latency across the reranking and mix-ranking stages by 13%. These results mark a paradigm shift in slate generation: moving beyond human-engineered heuristics toward a general learned policy creates a scalable path for further algorithmic optimization and compute-driven improvement.
Recommender systems are ubiquitous on the Web, making understandable explanations essential for users. Although large language models (LLMs) increasingly generate natural-language justifications, evaluating their factual grounding and personalized relevance remains challenging. Statement-level ranking offers an alternative by selecting concise, review-derived explanations, enabling fine-grained and reproducible evaluation using established ranking metrics. However, existing statement ranking methods largely rely on collaborative representations, while semantic enhancements often fail to outperform simple user-history popularity baselines. Moreover, directly adapting language models for personalized statement ranking remains largely unexplored. In this work, we introduce LMStaR (Language Models for Statement Ranking), a fully text-based framework that formulates personalized explanation ranking as dense retrieval. A query encoder represents the user's interaction history and target-item metadata, while a statement encoder maps candidate explanations into a shared embedding space. Both encoders are jointly adapted through multi-positive contrastive learning to learn personalized relevance. Experiments on Amazon Beauty, Amazon Toys, and Yelp show that LMStaR achieves the highest NDCG@1 across all three domains, outperforming CLER by 1.06--8.25 points (+11.8--65.9%). It also leads top-five ranking on both Amazon datasets, although CLER remains stronger on Yelp. Task-specific adaptation improves all five evaluated backbones and transfers across domains without target-domain fine-tuning. Counterfactual analyses further reveal the importance of user-specific histories. These findings highlight the potential of task-adapted language models for personalized explanation ranking.
Dense retrieval models are typically obtained by fine-tuning pretrained language models on supervised retrieval data, where high-quality annotations or interaction logs are often limited and reused across many epochs. This data-constrained regime differs from the abundant-data setting assumed by standard scaling laws, making it unclear how the best attainable retrieval performance changes with model size and unique fine-tuning data. We study this question through the minimum achievable evaluation loss (MAEL), defined as the lowest evaluation loss reached during training under fixed model-size and data-budget constraints. We first show that learning-rate choices strongly affect optimization trajectories but have little effect on the final attainable MAEL: after rescaling the optimization-step axis, learning curves collapse across learning rates, and CKA analysis shows that checkpoints reaching MAEL learn highly similar representations. We then fine-tune 35 BERT-family dense retrievers on MS MARCO subsets and find that MAEL is accurately described by a simple multiplicative power law over non-embedding parameters and unique samples, outperforming commonly used additive scaling formulations. We further derive a scaling law for the training step at which MAEL is reached and show that optimal weight decay follows an approximately invariant EMA timescale, yielding lower MAEL and stronger retrieval performance on both in-domain and out-of-domain benchmarks. Finally, combining loss and step scaling suggests that, in our observed regime, increasing pretrained model size is approximately 1.46 times as fine-tuning-compute-efficient as increasing unique in-domain data for reducing MAEL.
Content-category dashboards summarize production recommendation logs, although the serving policy selects the users, items, and contexts behind each mean. Using KuaiRand random-insertion impressions, we audit these summaries on the same 12,472 users and 5,486 videos in both logs. Standard means have absolute/centered mean absolute error (MAE) 0.1935/0.0491 and rank correlation 0.7833. User demeaning plus a random grand-mean anchor reduces absolute MAE by 80.0%, but centered MAE by only 21.0%, without reliably improving rank. A complete-support ledger allocates weighted squared error mainly to within-video residual (72.41%) and video mix (23.90%), with 3.65% from random support. In a frozen later test, early category calibration reaches MAE 0.0251 versus 0.0531 for scalar calibration; their direct difference is -0.0280 [-0.0296, -0.0172]. Decision evidence is cutoff-specific: the nominal top-10 regret change for row-weighted demeaning versus naive is -0.0062 [-0.0084, -0.0003], while top-5 and top-20 intervals include zero. The six-contrast family-wise top-10 interval [-0.0136, 0.0013] includes zero. An outcome-free disjoint randomized-reference split also leaves all six nominal contrasts covering zero. A known-truth stress test shows incomplete improvement under a mean-additive positive control and failure under item mix and weak overlap. All decompositions are descriptive and each result is conditional on its overlap cohort. KuaiRand intervals quantify user-cluster variation conditional on the realized catalog and calendar; we claim neither intrinsic category effects nor a universally valid correction.
Researchers struggle to find relevant datasets amid the growing volume of data across repositories and scholarly publications. Manual dataset retrieval is time-consuming and often fails to identify valuable resources that are not easily discoverable through direct links or popularity metrics. Dataset recommendation, which aims to identify datasets relevant to a paper, author, or query, has become essential for improving dataset discoverability and reuse. However, this task remains challenging because Scholarly Knowledge Graphs (SKGs) are large-scale, heterogeneous, dynamic, and incomplete. Although openly available SKGs organize scientific knowledge by linking scholarly outcomes, they often lack meaningful connections between datasets, leaving them isolated and underutilized. These graphs contain millions of evolving relations but still suffer from missing or inaccurate metadata, duplicate records, and weak dataset integration. This work presents the Multimodal Scholarly Attention Network (MM-SAN), a heterogeneous multimodal graph representation learning method for recommending datasets relevant to research publications within real-world SKGs. MM-SAN combines topology-aware and text-based embeddings, enabling effective recommendations even when textual metadata is limited or absent. It operates in transductive, semi-inductive, and fully inductive settings, adapting to evolving SKGs without requiring retraining. Experiments show that MM-SAN outperforms state-of-the-art methods, providing a robust and scalable solution for dataset recommendation.
In LLM-based recommendation, memory acts as a user-level agent in a memory-augmented backbone. Yet users and items remain asymmetric: only the user is given a structured profile; it enters a bandwidth-limited reranker prompt; and it is frozen after warmup. We recast LLM-based recommendation as an agentic-communication process in four stages: BUILD a self-description, TRANSMIT it along the collaborative graph, UNDERSTAND the other side, and SYNC it from the next interaction. This formulation reveals three unresolved communication challenges: the item side never produces a comparable self-description (one-sided BUILD); profile-level signals enter only the bandwidth-limited reranker prompt rather than retrieval, where they would not compete for context tokens (prompt-trapped UNDERSTAND); and the user's description is frozen after warmup with no principled update rule (absent SYNC). We instantiate the missing stages as HARP (Homophily-aware Agentic Recommendation via dual Profiles), a training-free, three-operator addition to a frozen LLM backbone: a rule-extracted symmetric item profile (zero extra LLM calls), a channel-decomposable profile-level homophily H(phi_u, phi_i) consumed at retrieval time rather than inside the reranker prompt, and a single-pass reflective update realising in-context profile evolution without any gradient step. On four public benchmarks (Yelp, Books, MovieTV, Goodreads), HARP lifts Hit@1 over the strongest published baseline by 6.9%-24.0%, with family-wise Holm-corrected p < 1e-9 on every dataset. Ablations and a per-case attribution study show that the operators contribute complementary gains, with the retrieval-side gate as the main source; we also report the trade-off introduced by the reflective pass. We formulate agentic communication and provide a training-free instantiation. Code is available at https://github.com/Jove-Zhen/HARP.
Next-visit diagnosis prediction forecasts the diagnoses documented at a patient's next encounter, turning retrospective electronic health records into prospective clinical prediction. However, conventional aggregate evaluation obscures its multifaceted nature: distributionally, models must handle variable-size diagnosis sets and long-tailed labels; temporally, they must distinguish persistent, newly recorded, and recurrent diagnoses; and both vary with observed history depth. We introduce an evaluation framework along these dimensions and apply it to rule-based heuristics, LLM-based prompting, retrieval, memory, and multi-agent approaches. This reveals that repeating the last visit outperforms most LLM-based approaches overall despite predicting no new or recurrent diagnoses, and that retrieval and memory exhibit distinct profiles whose relative strengths vary across temporal transitions and backbones. Motivated by this, we propose LONGINUS, a retrieval-memory multi-agent adjudication framework combining evidence retrieved from patient records with structured memory of the patient trajectory. Using cross-view agreement as a reliability heuristic, it preserves shared predictions and adjudicates disagreements without revealing their source. Across MIMIC-III and MIMIC-IV with four LLM backbones, LONGINUS achieves the best transition-specific performance among LLM-based methods in nearly all settings while remaining competitive with them on aggregate metrics.
Answering geographic queries over Earth observation (EO) archives requires solving a joint retrieval-and-derivation problem under strict spatiotemporal constraints. Existing LLM agents fail to ground observation items or suffer from search-space explosion. We propose OSGR, a two-stage framework that factorizes this search into offline symbolic concept deduction and online Relation-Conditioned Grounding. OSGR leverages an offline Observation-Support Graph to deduce valid target-to-result support paths at the concept level, and applies spatiotemporal compatibility kernels to instantiate grounded observation-item sets. On Earth-Bench and DisasterM3-Reason, OSGR improves GPT-4o accuracy from 49.3\% to 63.7\% while reducing the execution-step ratio from 2.9 to 2.5.
Generative recommendation is gaining traction in production, retrieving items by generating their identifiers one token at a time. The generated sequences must identify relevant catalog items, so constrained decoding couples relevance scoring with validity filtering. Because most possible sequences are invalid, scoring before filtering wastes computation on candidates that will be discarded. Our evaluation shows that filtering first can avoid this work with negligible impact on recommendation quality. We present ZeroGR, a filtering-first constrained decoding system for generative recommendation. ZeroGR uses a step-sliced filtering index to reduce GPU memory use by tailoring data representations to each decoding step. Its execution engine adopts shared-slot execution to accelerate data exchange across stages and hybrid projection to adapt computation to workload patterns, thereby reducing execution overhead. During index refreshes, ZeroGR applies layout-guided streaming construction to speed up updates and limit serving interference. Extensive evaluation shows that ZeroGR accelerates the filtering and scoring pipeline by up to $9.9\times$ over representative baselines while reducing GPU index memory use and accelerating index reconstruction.
Web systems infer user preferences from behavior, ratings, text, and LLM-generated profiles, yet explaining a proxy may reveal its measurement process rather than the preference of interest. We develop an audit framework distinguishing point identification, partial identification, and abstention relative to an explicit reference distribution. Under feature-dependent, arbitrarily signed proxy bias, a strict majority of uncontaminated proxy attributions lets the coordinatewise median recover latent-attribution support and direction and, with externally justified positive scales, magnitude. We prove this threshold sharp using response-valid indistinguishable worlds, characterize sharp minimax error and exact compatible-target sets under approximate cleanliness and scale uncertainty, and propagate clean-proxy attribution error. A frozen known-truth benchmark spans two response families, four proxy counts, all contamination counts, three bias strengths, and eleven methods, with a focal comparison confirmed on twenty unseen seeds. Under correct budgets, a direct attribution-tensor audit achieved 100% coverage, 95.40%-97.05% active-driver sign decisions, and no false signs. In a temporally separated Amazon study, median aggregation reduced comparator-relative conditional sign error for all four targets; six of eight prespecified target-metric contrasts had Bonferroni intervals excluding zero. A controlled real-covariate perturbation favored the median on the prespecified focal metric for all four targets. Compatible sets showed that the current proxy configuration rarely licensed strict-sign conclusions. Weak comparator calibration limits the Amazon evidence to fitted-surface alignment rather than latent-preference recovery.
Composed Image Retrieval (CIR) aims to retrieve a target image given a reference image and a modification text. Although recent approaches leveraging Multimodal Large Language Models (MLLMs) have shown significant progress, we argue that the MLLM's compositional understanding remains underutilized. This stems from directly mapping composed queries into a retrieval embedding via contrastive supervision alone, without explicitly resolving the intended visual modifications. In this work, we introduce Factorized MLLM Embeddings (FactEmbed), which decouples compositional query comprehension from building discriminative embeddings. Rather than directly projecting the composed query into a single embedding without internal reasoning, FactEmbed employs a set of factorized latents that simulate the intended visual modifications via an intermediate target imagination objective. This process encourages the latents to ground into target semantics before forming discriminative representations for retrieval. Subsequently, to propagate these factorized representations throughout the model for retrieval, we optimize a factor-level contrastive loss while distilling the decomposed representations into a deeper attention layer of the MLLM. Finally, for efficient inference, we retain the [EOS] token to establish a global-level embedding on top of these factorized latents, while we exploit the mean-pooled factorized latents to rerank a short candidate list at negligible cost, as both are produced by the same forward pass. Extensive evaluations demonstrate that FactEmbed outperforms prior works across CIR benchmarks.