
ALT Personalized content recommendation on large marketplace websites must reconcile heterogeneous candidate content (e.g., customer stories, product pages, industry pages, and learning content) with sparse, fast-changing user behavior, placing increasing demands on the retrieval stage to represent user interest well, not just the downstream ranking stage. We present LSER (LLM Session Embedding Retrieval), a production content recommendation system for autodesk.com that derives session- and history-level embeddings from a pretrained large language model (LLM) and adds them as auxiliary user-side features in the retrieval-stage user tower alongside existing learned sequence encoders. To keep user-side embeddings fresh under real-time behavior, we build a nearline serving path that asynchronously refreshes sequential and LLM-derived user vectors for our main content recommendation server. We validate the incremental value of the LLM features through an ablation study across our full multi-co

ALT In recent decades, galleries, libraries, archives, and museums (GLAM institutions) have digitised vast amounts of cultural heritage material, including historical imagery. A major challenge in making these collections publicly accessible is enabling effective search and discovery. Many historical images contain incomplete, sparse, or no descriptive metadata. Consequently, machine learning (ML) methods are increasingly being used to generate or enrich metadata to improve the searchability of such collections. However, artificial intelligence (AI)-based systems can reproduce or amplify existing biases, particularly when applied to historical imagery. Factors such as differences between the data used to train the models and the historical images on which they are applied, degraded image quality, monochrome photographs, and historically biased or incomplete descriptions can reduce model performance and introduce bias into search and retrieval results. In this paper, we examine different fo

ALT Current research agents are moving beyond academic search toward literature synthesis and experiment planning. These workflows increasingly rely on literature retrieval systems, which serve as standalone systems or as components of research agents. However, existing scientific literature retrieval benchmarks mainly evaluate these methods using topical relevance or citation links as supervision, which cannot capture which prior methods are used for empirical comparison. To address this gap, we introduce Lineage, a benchmark for empirical lineage retrieval in AI scientific literature. An empirical lineage relation links a prior paper to a source paper that explicitly uses its method as an empirical comparison baseline. These relations form lineage graphs whose multi-hop neighborhoods provide retrieval ground truth for structural evaluation. To construct these graphs at scale, we develop Lineage-Constructor, which converts heterogeneous paper records and full-paper empirical baseline evid

ALT Continuous generative retrievers powered by flow matching and diffusion sample multiple queries in item embedding space for parallel approximate nearest-neighbor (ANN) search under fixed retrieval budgets. However, optimizing distribution diversity or target-level interest separation ignores set-level retrieval geometry: sampled queries collapse onto overlapping ANN neighborhoods, while discrete top-K lookup and single-target supervision block end-to-end coverage optimization. We introduce CACO (Coverage-Aware Coordinated Optimization), a lightweight self-supervised post-training framework that coordinates queries from frozen flow-matching and diffusion retrievers toward complementary item coverage with negligible additional inference latency. CACO uses empirical retrieval frequencies across sampled reference queries as a relevance proxy, formulating a label-free coverage objective that improves recall across all 30 backbone-dataset settings in offline greedy selection experiments. To

ALT Resolving entity conflicts is a critical step in Pipeline-based NLU architectures, especially when multiple extractors provide overlapping or conflicting outputs. Traditional rule-based entity conflict resolution methods often lack flexibility and are unscalable. In this work, we aim to study if the broad knowledge of LLMs can be leveraged to resolve entity conflicts in domain specific NLU pipelines. We present an experimental comparison between a traditional rule-based conflict resolution method, and two LLM-based approaches: a multi-shot prompting method using the Llama3.1-8B-Instruct model and a LoRA fine-tuned Llama3.1-8B-Instruct model. The evaluation was conducted under two setups: (1) a broad assessment of LLM's capability to match existing NLU pipeline output, and (2) a more controlled scenario assessing the LLM accuracy on human-labeled data. Results show that while multi-shot prompting offers ease of adaptation and rapid behavioral changes, the fine-tuned model delivers super

ALT Discovery of relevant datasets within data lakes remains a key challenge. Union and join search are effective techniques for discovering data when augmenting an existing table. In exploratory search, instead, keyword search is commonly adopted to find a relatively small set of highly relevant tables matching some search terms. In this context, semantic search is a new technique that exploits the presence of named entities in exemplar tuples used as queries and in target tables. Hence, keyword search and entity search are able to find a diverse set of tables. In this work, we propose Thetis++, which combines keyword and semantic table search to discover relevant tables that combine the semantic signal with the presence of exact keyword matches. Incorporating multiple paradigms and sources of information allow us to combine the best of multiple worlds. Therefore, we study many alternative techniques for combining rankings and evaluate these for the table search task in data lakes. We exp

ALT Long user behavior sequences are among the strongest signals for click-through rate (CTR) prediction. Cross-attention makes these sequences practical for industrial ranking by keeping compute linear in sequence length. Yet the gains from stacking cross-attention layers taper off: on Taobao search, growing a cross-attention stack from 16 to 28 layers adds only 0.02 percentage points (pt) of AUC. We argue that the key-value side is the bottleneck. Each layer projects the same behavior sequence into keys and values, so adding depth gives the model more reads of a fixed sequence. We propose DNA (Dual-stream Nexus Attention), which lets the sequence itself evolve with depth. At each layer, the request representation (the search query, user features and scene) first reads the sequence. Each behavior token then attends to the updated request representation and uses what it reads to update itself, so the next layer reads a sequence shaped by the current request. The update runs once per reques

ALT Text-based generative recommenders can go beyond ranking existing items by directly generating item titles or other textual representations. In the news domain, this raises a societal concern that models may repeatedly surface already popular articles and thereby narrow the diversity of information to which users are exposed. We investigate training-data memorization as one mechanism behind this concentration. Using approximately 20 million browsing sessions from a Japanese news service, we continually trained an instruction-tuned Llama 3 8B model to generate the next article title in a session. The generated-title distribution was more concentrated and less diverse than the distribution of titles observed in the browsing logs. Frequently repeated target titles exhibited greater verbatim and approximate memorization, and more strongly memorized titles were generated in response to a wider range of prompts. These findings support a pathway in which popularity in behavioral logs becomes

ALT A recommendation page is consumed as a whole: the items on it compete for attention and complement one another, so the best page is not simply the individually best items placed side by side. We study whole-page optimization, which selects and orders the items on a page jointly to maximize the score of the page. The number of possible pages grows combinatorially, so good pages must be generated rather than enumerated. Some generative re-ranking methods use the same page-level training signal for every choice in a sampled page. Others imitate target pages built before training, which remain largely fixed as the generator improves. We address both limitations with one local page-level objective, which repeatedly moves the current generator toward pages with higher score. This objective splits exactly into one training target per position, and each target compares candidates at the same partial page by the scores of the complete pages they lead to. Every position therefore learns to choos

ALT Generative retrieval (GR) for sequential recommendation compresses each item into a short semantic ID and trains an autoregressive decoder to generate next-item codes from a user history. While appealing for cold-start handling and storage, au- toregressive GR suffers a structural scaling limit: typically, longer semantic IDs reduce quantization error in principle but degrade ranking quality and quadratic memory usage against semantic ID length. We propose Late Quantization (LQ), a non-autoregressive paradigm that confines discretization to the output (item) side of the pipeline. The encoder and scoring rule operate in continuous space, while items are stored and matched as semantic IDs produced by an end-to-end-trained residual quantizer. This architectural shift permits an additive error decomposition and a pairwise ordering theorem that jointly bound the quantization- induced ranking drop, a structural guarantee absent in prior semantic-ID recommenders. Across four public benchmarks

ALT The correspondence of Heinrich Bullinger (1504-1575), Swiss Reformer and Antistes of Zurich, documents the political and religious debates of the Reformation, including extensive appeals to the Church Fathers as authoritative sources. Detecting such references computationally remains largely unexplored. We investigate dense and sparse retrieval for patristic reference detection, and metadata-informed topic modelling as a proxy for the topics and ideas discussed. Prior work on Latin intertextuality has fine-tuned dense retrievers for their respective target domain, and we test how well these models transfer to Bullinger's correspondence. We compare lexical (BM25), dense (bi-encoders), and hybrid methods on explicit and implicit references. A Latin encoder fine-tuned on biblical reuse performs best among the bi-encoders (82%, Recall@20 on explicit and 16% on implicit references), while BM25 outperforms dense approaches on explicit citations (94%, R@20). Because BM25 and dense retrieval r

ALT Large Language Models (LLMs) have demonstrated strong capabilities in open-domain question answering, but often struggle with factual accuracy and multi-hop reasoning due to the incompleteness of the training corpora. A promising solution is Knowledge Graph Retrieval-Augmented Generation (KG-RAG), which supplements LLMs with structured knowledge retrieved from external knowledge graphs (KGs). However, existing KG-RAG methods either rely on large-scale language models (e.g., ChatGPT) to guide the retrieval process, which leads to high computational costs, or suffer from limited retrieval quality when using lightweight language models, particularly under multi-hop scenarios. We propose LSP-RAG (Latent Scoring Perspectives for Retrieval-Augmented Generation over Knowledge Graphs), a novel KG-RAG framework that enhances hop-wise KG knowledge retrieval through multi-perspective prompted scoring. Each latent scoring perspective is guided by a combination of two types of soft prompts: a persp

ALT Large language model (LLM) judges have become a common way to create large-scale relevance judgments for retrieval evaluation. However, each judgment requires a potentially expensive call to a frontier model, and the judgments are hard to interpret and may not be reproducible. In this paper, we ask whether an LLM can write a readable program for each topic that replaces its per-document judgments. To address these issues, we propose ExecJudges, a method for generating fast, inspectable and scalable judges for automatic relevance labeling. Specifically, ExecJudges uses a frontier LLM to turn assessment guidelines into executable programs, which rely only on lexical matching, standard NLP libraries like parsers, and pool statistics. An ExecJudges-L variant also uses examples of human labels if available. We evaluate ExecJudges on seven TREC collections by comparing the produced labels with trained NIST assessors, frontier LLM judges, and recently proposed per-topic classifiers. Our resul

ALT Accurate click-through rate (CTR) prediction depends on user representations that capture both persistent preferences and recent intent, yet processing extensive behavioral histories for every ranking request is costly under production compute and latency constraints. We present X-split ARGUS, a framework for ad ranking whose history compressor and online ranker are trained end to end. The compressor encodes past interactions into a compact set of tokens that are computed offline and reused across candidates, while the online ranker scores ads through cross-attention to these tokens and representations of recent events. Training combines an autoregressive loss for click feedback with distillation of candidate-specific click logits from a larger teacher that scores each candidate using the uncompressed history. The number of compressed history tokens and the length of recent history can be adjusted to fit serving compute and latency budgets. Offline evaluation shows a 0.81% relative red

ALT Modern industrial recommender systems use multi-stage cascadesโretrieval, pre-ranking, ranking, and re-rankingโto narrow millions of candidates to a personalized list. While each stage deploys sophisticated models, pipeline-level decisions on channel activation, quotas, model selection, and score fusion remain engineer-tuned and user-agnostic within each scenario. This report presents MARS (Multi-stage Agentic Recommendation System), extending personalization from individual-stage predictions to full pipeline orchestration. A near-line agent built on fine-tuned Qwen3-0.6B generates per-user configurationsโa seven-dimensional action covering pipeline mode, retrieval quotas, model selection, fusion weights, and reranker settingsโfrom a constrained action space. Configurations are validated, cached, and consumed asynchronously; the cascade never waits for LLM inference. Schema checks, fallbacks, and a kill switch ensure serving safety. The policy is trained in three stages: Confidence-awa

ALT Multimodal generative recommendation with discrete semantic identifiers (SIDs) requires distinguishing historical modalities that merely alter token predictions from those that genuinely reduce catalog-level ambiguity. When audiovisual, visual, or textual streams are missing, degraded, or semantically conflicting, uniform fusion and unconstrained token gates can propagate misleading content across iterative denoising steps. We propose CONCORD-Rec, a reliability-conditioned discrete diffusion recommender supervised by counterfactual catalog-partition utility. Given a fixed tokenizer and a partially masked SID state, CONCORD-Rec evaluates out-of-fold teacher coalitions over the exact feasible catalog support and aggregates item posteriors across the target code partition at each unresolved position. Rather than distilling raw token probabilities directly, the resulting counterfactual loss differences supervise an availability-masked tokenwise gate that allocates bounded modality residual

ALT Criminal court records are often fragmented, noisy, and siloed, making it challenging for both legal experts and community stakeholders to access and analyze court data to identify trends and provide public oversight. In this paper, we introduce a system for data ingestion, harmonization, and dashboard visualization designed for New Orleans criminal court proceedings data, aimed at both expert and lay audiences. Our system collects and harmonizes disparate data, generates case trajectories, and provides intuitive visualization techniques based on community feedback to analyze demographics, temporal trends, and specific case events. We demonstrate the system through an exploratory analysis of recent short-term trends and policy impacts within New Orleans, illustrating how it can support the investigation of anomalies, policy changes, and procedural irregularities. We also present feedback from a small formative study of community stakeholders who have helped co-design and use the system

ALT Conversational meme retrieval supports response selection in online communication, where relevance depends on both contextual meaning and affective expression. We study this task using two language-specific collections from Douban and Imgflip, containing 98,506 meme records across 5,583 discussion topics. Thread membership supplies weak relevance labels, while likes provide implicit ranking feedback. Automatically extracted sentiment, facial-expression indicators, and cross-modal consistency features support an exploratory analysis of engagement. We propose CARE, a Conversational Affective Response Retrieval framework that combines an emotion-enhanced CLIP retriever with a Jina-based reranker. The retriever incorporates free-form emotion descriptors through gated fusion and learns from multiple positive examples per query. The reranker combines queryโimage interactions with emotion representations and is trained using engagement-based ranking supervision. Under the reported topic-held-

ALT In this work we propose a lightweight, interpretable approach that estimates item-item music similarity from publicly available metadata. Our hybrid regression model combines content-based filtering with the classic item-based collaborative filtering. We test generalization via a zero-shot evaluation on Latin American music, a genre for which co-occurrence data is scarce. Accurately predicting similarity for these unseen, sparse pairs shows that our approach is robust to the regional gap affecting many existing systems, without the audio signal processing or the computational budget that deep multimodal approaches would require.

ALT Generative recommenders decode the semantic ID of a user's next item with a language model, and post-training them with reinforcement learning (RL) after supervised fine-tuning (SFT) has become standard practice. The gains are mostly reported on the first ten results, yet in deployment such models hand hundreds of candidates to a downstream ranker or reward model. We show that RL post-training has a hidden cost: it improves accuracy at the top of the list (hit@K for small K) but, on all five datasets we study, lowers the recall of the candidate list (hit@K for K in the hundreds). The model becomes a better ranker and a worse retriever. Controlled experiments trace the loss to the negative advantages that RL assigns to the candidates it samples, the rollout. The root cause is that only one next item is logged per user, so RL treats every other candidate as wrong. Yet these candidates are the model's most plausible alternatives, and often the next items of other users with similar histor
