arXiv subjects

Shubham Chatterjee

Publications and source records attributed to Shubham Chatterjee.

At least 19 recordsLinked to original sources

arXiv2026

Assessing the Downstream Utility of Evidence-Aware Retrieval in RAG

Retrieval evaluation for retrieval-augmented generation (RAG) is increasingly designed around whether retrieved passages contain evidence that can support generation, rather than topical relevance alone. We study whether this closer alignment with downstream evidence needs also makes retrieval evaluation more useful for the decisions built from it. Across five retrieval benchmarks and an end-to-end TREC RAG 2025 setting, we examine an answer-support signal in four roles: comparing retrievers, guiding retrieval training and system selection, predicting downstream answer quality, and filtering the evidence supplied to a generator. The signal changes retrieval rankings, but its downstream value is not uniform. It does not reliably improve retriever training; the benefit of using it for system selection depends on how the generator is instructed to use the retrieved evidence; and retrieval scores based on it do not robustly predict answer quality on unseen topics. In a direct evidence intervention, human annotators confirm that filtering preferentially preserves passages containing useful answer evidence, yet different answer evaluators reach different conclusions about whether the resulting answers improve. These results show that making retrieval evaluation more closely reflect the evidence needed for generation does not by itself make every downstream use of that evaluation more reliable. RAG evaluation methods should therefore be assessed with respect to the particular comparisons, decisions, and conclusions they are intended to support.

cs.IR↗

arXiv2026

When Do Anchor-Based Pointwise LLM Rerankers Help? Retriever Quality, Statistical Scope, and Anchor Design

Anchor-based pointwise LLM reranking scores each candidate against a shared reference passage to recover cross-document context at pointwise cost. We study when this actually helps, using GCCP/PAGC as a representative method. Our study is reproduction-first. We use reproduction as a starting point for a controlled component-level stress test of anchor-based pointwise reranking. Our initial reimplementation, based only on the paper text, achieves 0.24 nDCG@10 instead of the reported 0.66, revealing that several undocumented implementation details are necessary to reproduce the method. After identifying and recovering eight such details, we reproduce the reported results within 1.6% and use the validated implementation for controlled analysis. We find that the core contrastive scoring idea is robust under rigorous statistical correction. However, two design choices held fixed in the original paper are less reliable. First, we find that combining the contrastive score with the standard pointwise relevance score helps when the first-stage retriever is BM25, but gives little or no benefit when the first-stage retriever is a stronger dense model such as E5. Second, the paper's more complex method for constructing the anchor is unnecessary. A much simpler anchor, built by interleaving the top-ranked sentences, matches or outperforms it across datasets. These findings are consistent across different LLM backbones, including a 4-bit quantized 72B model. Overall, anchor-based pointwise reranking is effective, but its gains come mainly from contrastive scoring rather than from the more complex aggregation and anchor-construction choices, and they appear under narrower conditions than the original evaluation suggests.

cs.IR↗

arXiv2026

Bridge Evidence: Static Retrieval Utility Does Not Predict Causal Utility in Multi-Step Agentic Search

Retrieval systems are trained and evaluated on a static idea of usefulness: hand a document and a question to a reader model, see whether the answer improves, and score the document accordingly. The idea holds up when a document is read on its own. It breaks when a language model works as a search agent, issuing several queries and reasoning across turns, because a document can matter for what it lets the agent do next rather than for what it says about the current question. We measure that gap rather than argue it. Using a ReAct style agent over HotpotQA, we replay 1000 development questions and, for every document the agent read, delete it and re-run the rest of the trajectory from that point. Comparing the original run against its counterfactual gives a Counterfactual Trajectory Utility (CTU) score from three deltas: final answer quality, next query retrieval quality, and turn count. Crossing CTU against Static RAG Utility (SRU) over 23,322 document observations, the two are close to statistically independent (Spearman rho = -0.026). Roughly a third of the documents the agent reads are causally load bearing while looking useless to a static reader; we call these bridge documents. The pattern survives when the reader based axis is swapped for a BM25 and cross encoder proxy, giving a bridge cell of 27.2% on an evenly spread axis. A second experiment pins down the mechanism. Using the Observable Entity Relevance (OER) measure from prior work, entities that discriminate relevant from non-relevant candidates appear in the agent's next query 4.02 times more often than entities found only in non-relevant documents (6.1% vs 1.5%, n = 227,139). A bridge document earns its keep by handing the agent a discriminative entity that redirects the search. Static relevance and causal usefulness are different quantities in agentic retrieval, and optimizing the first does not deliver the second.

cs.IR↗

arXiv2026

Entity Labels Are Not Entity Signals: A Framework for Observable Relevance in Document Re-Ranking

Entity-aware document retrieval uses query-associated entities as ranking signals, assuming that semantically relevant entities are also useful retrieval signals. We show this assumption is insufficient- and explain why. Unlike terms, which are ground-truth observations, entity links are hypotheses produced by an imperfect linker: an entity can be topically central yet provide no discriminative signal if the linker fires indiscriminately across relevant and non-relevant documents. We formalize this as a distinction between Conceptual Entity Relevance (CER)- whether an entity is topically related to a query- and Observable Entity Relevance (OER)- whether its observed presence in a collection discriminates relevant from non-relevant documents. Across four collections and annotation sources including human entity judgments, CER and OER exhibit near-chance agreement ($\kappa \approx 0$), while OER operationalizations agree substantially ($\kappa \approx 0.5$), confirming CER as the systematic outlier. CER-based supervision selects topically plausible but weakly discriminative entities, pruning fewer than 4% of non-relevant documents on some collections. Aligning supervision with OER improves non-relevant pruning by up to 10x and open-world MAP by 0.051 over BM25. Our findings motivate a shift from conceptual to observable notions of entity relevance in entity-aware retrieval.

cs.IR↗

arXiv2026

Reproduction Beyond Benchmarks: ConstBERT and ColBERT-v2 Across Backends and Query Distributions

Reproducibility must validate architectural robustness, not just numerical accuracy. We evaluate ColBERT-v2 and ConstBERT across five dimensions, finding that while ConstBERT reproduces within 0.05% MRR@10 on MS-MARCO, both models show a drop of 86-97% on long, narrative queries (TREC ToT 2025). Ablations prove this failure is architectural: performance plateaus at 20 words because the MaxSim operator's uniform token weighting cannot distinguish signal from filler noise. Furthermore, undocumented backend parameters create an 8-point gap due to ConstBERT's sparse centroid coverage, and fine-tuning with 3x more data actually degrades performance by up to 29%. We conclude that architectural constraints in multi-vector retrieval cannot be overcome by adaptation alone. Code: https://github.com/utshabkg/multi-vector-reproducibility.

cs.IR↗

arXiv2026

Entities as Retrieval Signals: A Systematic Study of Coverage, Supervision, and Evaluation in Entity-Oriented Ranking

Entity-oriented retrieval assumes that relevant documents exhibit query-relevant entities, yet evaluations report conflicting results. We show this inconsistency stems not from model failure, but from evaluation. On TREC Robust04, we evaluate six neural rerankers and 437 unsupervised configurations against BM25. Across 443 systems, none improves MAP by more than 0.05 under open-world evaluation over the full candidate set, despite strong gains under entity-restricted settings. The best configuration matches the official Robust04 best system and outperforms most neural rerankers, indicating that the architecture is not the limiting factor. Instead, the bottleneck is the entity channel: even under idealized selection, entity signals cover only 19.7\% of relevant documents, and no method achieves both high coverage and discrimination. We explain this via a distinction between Conceptual Entity Relevance (CER) -- semantic relatedness -- and Observable Entity Relevance (OER) -- corpus-grounded discriminativeness under a given linker. All supervision strategies operate at the CER level and ignore the linking environment, leading to signals that are semantically valid but not discriminative. Improving supervision therefore does not recover open-world performance: stronger signals reduce coverage without improving effectiveness. Conditional and open-world evaluation answer different questions: exploiting entity evidence versus improving retrieval under realistic linking, but are often conflated. Progress requires datasets with entity-level discriminativeness and evaluation that reports both coverage and effectiveness. Until then, conditional gains do not imply open-world effectiveness, and open-world failures do not invalidate entity-based models.

cs.IR↗

arXiv2026

Interplay: Training Independent Simulators for Reference-Free Conversational Recommendation

Training conversational recommender systems (CRS) requires extensive dialogue data, which is challenging to collect at scale. To address this, researchers have used simulated user-recommender conversations. Traditional simulation approaches often utilize a single large language model (LLM) that generates entire conversations with prior knowledge of the target items, leading to scripted and artificial dialogues. We propose a reference-free simulation framework that trains two independent LLMs, one as the user and one as the conversational recommender. These models interact in real-time without access to predetermined target items, but preference summaries and target attributes, enabling the recommender to genuinely infer user preferences through dialogue. This approach produces more realistic and diverse conversations that closely mirror authentic human-AI interactions. Our reference-free simulators match or exceed existing methods in quality, while offering a scalable solution for generating high-quality conversational recommendation data without constraining conversations to pre-defined target items. We conduct both quantitative and human evaluations to confirm the effectiveness of our reference-free approach.

cs.AI↗

arXiv2026

Single-Turn LLM Reformulation Powered Multi-Stage Hybrid Re-Ranking for Tip-of-the-Tongue Known-Item Retrieval

Retrieving known items from vague descriptions, Tip-of-the-Tongue (ToT) retrieval, remains a significant challenge. We propose using a single call to a generic 8B-parameter LLM for query reformulation, bridging the gap between ill-formed ToT queries and specific information needs. This method is particularly effective where standard Pseudo-Relevance Feedback fails due to poor initial recall. Crucially, our LLM is not fine-tuned for ToT or specific domains, demonstrating that gains stem from our prompting strategy rather than model specialization. Rewritten queries feed a multi-stage pipeline: sparse retrieval (BM25), dense/late-interaction reranking (Contriever, E5-large-v2, ColBERTv2), monoT5 cross-encoding, and list-wise reranking (Qwen 2.5 72B). Experiments on 2025 TREC-ToT datasets show that while raw queries yield poor performance, our lightweight pre-retrieval transformation improves Recall by 20.61%. Subsequent reranking improves nDCG@10 by 33.88%, MRR by 29.92%, and MAP@10 by 29.98%, offering a cost-effective intervention that unlocks the potential of downstream rankers. Code and data: https://github.com/debayan1405/TREC-TOT-2025

cs.IR↗

arXiv2025

QDER: Query-Specific Document and Entity Representations for Multi-Vector Document Re-Ranking

Neural IR has advanced through two distinct paths: entity-oriented approaches leveraging knowledge graphs and multi-vector models capturing fine-grained semantics. We introduce QDER, a neural re-ranking model that unifies these approaches by integrating knowledge graph semantics into a multi-vector model. QDER's key innovation lies in its modeling of query-document relationships: rather than computing similarity scores on aggregated embeddings, we maintain individual token and entity representations throughout the ranking process, performing aggregation only at the final scoring stage - an approach we call "late aggregation." We first transform these fine-grained representations through learned attention patterns, then apply carefully chosen mathematical operations for precise matches. Experiments across five standard benchmarks show that QDER achieves significant performance gains, with improvements of 36% in nDCG@20 over the strongest baseline on TREC Robust 2004 and similar improvements on other datasets. QDER particularly excels on difficult queries, achieving an nDCG@20 of 0.70 where traditional approaches fail completely (nDCG@20 = 0.0), setting a foundation for future work in entity-aware retrieval.

cs.IR↗

arXiv2025

REGENT: Relevance-Guided Attention for Entity-Aware Multi-Vector Neural Re-Ranking

Current neural re-rankers often struggle with complex information needs and long, content-rich documents. The fundamental issue is not computational--it is intelligent content selection: identifying what matters in lengthy, multi-faceted texts. While humans naturally anchor their understanding around key entities and concepts, neural models process text within rigid token windows, treating all interactions as equally important and missing critical semantic signals. We introduce REGENT, a neural re-ranking model that mimics human-like understanding by using entities as a "semantic skeleton" to guide attention. REGENT integrates relevance guidance directly into the attention mechanism, combining fine-grained lexical matching with high-level semantic reasoning. This relevance-guided attention enables the model to focus on conceptually important content while maintaining sensitivity to precise term matches. REGENT achieves new state-of-the-art performance in three challenging datasets, providing up to 108% improvement over BM25 and consistently outperforming strong baselines including ColBERT and RankVicuna. To our knowledge, this is the first work to successfully integrate entity semantics directly into neural attention, establishing a new paradigm for entity-aware information retrieval.

cs.IR↗

arXiv2024

DyVo: Dynamic Vocabularies for Learned Sparse Retrieval with Entities

Learned Sparse Retrieval (LSR) models use vocabularies from pre-trained transformers, which often split entities into nonsensical fragments. Splitting entities can reduce retrieval accuracy and limits the model's ability to incorporate up-to-date world knowledge not included in the training data. In this work, we enhance the LSR vocabulary with Wikipedia concepts and entities, enabling the model to resolve ambiguities more effectively and stay current with evolving knowledge. Central to our approach is a Dynamic Vocabulary (DyVo) head, which leverages existing entity embeddings and an entity retrieval component that identifies entities relevant to a query or document. We use the DyVo head to generate entity weights, which are then merged with word piece weights to create joint representations for efficient indexing and retrieval using an inverted index. In experiments across three entity-rich document ranking datasets, the resulting DyVo model substantially outperforms state-of-the-art baselines.

cs.IR↗

arXiv2024

Doing Personal LAPS: LLM-Augmented Dialogue Construction for Personalized Multi-Session Conversational Search

The future of conversational agents will provide users with personalized information responses. However, a significant challenge in developing models is the lack of large-scale dialogue datasets that span multiple sessions and reflect real-world user preferences. Previous approaches rely on experts in a wizard-of-oz setup that is difficult to scale, particularly for personalized tasks. Our method, LAPS, addresses this by using large language models (LLMs) to guide a single human worker in generating personalized dialogues. This method has proven to speed up the creation process and improve quality. LAPS can collect large-scale, human-written, multi-session, and multi-domain conversations, including extracting user preferences. When compared to existing datasets, LAPS-produced conversations are as natural and diverse as expert-created ones, which stays in contrast with fully synthetic methods. The collected dataset is suited to train preference extraction and personalized response generation. Our results show that responses generated explicitly using extracted preferences better match user's actual preferences, highlighting the value of using extracted preferences over simple dialogue history. Overall, LAPS introduces a new method to leverage LLMs to create realistic personalized conversational data more efficiently and effectively than previous methods.

cs.IR↗

arXiv2024

TREC iKAT 2023: A Test Collection for Evaluating Conversational and Interactive Knowledge Assistants

Conversational information seeking has evolved rapidly in the last few years with the development of Large Language Models (LLMs), providing the basis for interpreting and responding in a naturalistic manner to user requests. The extended TREC Interactive Knowledge Assistance Track (iKAT) collection aims to enable researchers to test and evaluate their Conversational Search Agents (CSA). The collection contains a set of 36 personalized dialogues over 20 different topics each coupled with a Personal Text Knowledge Base (PTKB) that defines the bespoke user personas. A total of 344 turns with approximately 26,000 passages are provided as assessments on relevance, as well as additional assessments on generated responses over four key dimensions: relevance, completeness, groundedness, and naturalness. The collection challenges CSA to efficiently navigate diverse personal contexts, elicit pertinent persona information, and employ context for relevant conversations. The integration of a PTKB and the emphasis on decisional search tasks contribute to the uniqueness of this test collection, making it an essential benchmark for advancing research in conversational and interactive knowledge assistants.

cs.IR↗

arXiv2024

DREQ: Document Re-Ranking Using Entity-based Query Understanding

While entity-oriented neural IR models have advanced significantly, they often overlook a key nuance: the varying degrees of influence individual entities within a document have on its overall relevance. Addressing this gap, we present DREQ, an entity-oriented dense document re-ranking model. Uniquely, we emphasize the query-relevant entities within a document's representation while simultaneously attenuating the less relevant ones, thus obtaining a query-specific entity-centric document representation. We then combine this entity-centric document representation with the text-centric representation of the document to obtain a "hybrid" representation of the document. We learn a relevance score for the document using this hybrid representation. Using four large-scale benchmarks, we show that DREQ outperforms state-of-the-art neural and non-neural re-ranking methods, highlighting the effectiveness of our entity-oriented representation approach.

cs.IR↗

arXiv2024

TREC iKAT 2023: The Interactive Knowledge Assistance Track Overview

Conversational Information Seeking has evolved rapidly in the last few years with the development of Large Language Models providing the basis for interpreting and responding in a naturalistic manner to user requests. iKAT emphasizes the creation and research of conversational search agents that adapt responses based on the user's prior interactions and present context. This means that the same question might yield varied answers, contingent on the user's profile and preferences. The challenge lies in enabling Conversational Search Agents (CSA) to incorporate personalized context to effectively guide users through the relevant information to them. iKAT's first year attracted seven teams and a total of 24 runs. Most of the runs leveraged Large Language Models (LLMs) in their pipelines, with a few focusing on a generate-then-retrieve approach.

cs.IR↗

arXiv2023

Adaptive Latent Entity Expansion for Document Retrieval

Despite considerable progress in neural relevance ranking techniques, search engines still struggle to process complex queries effectively - both in terms of precision and recall. Sparse and dense Pseudo-Relevance Feedback (PRF) approaches have the potential to overcome limitations in recall, but are only effective with high precision in the top ranks. In this work, we tackle the problem of search over complex queries using three complementary techniques. First, we demonstrate that applying a strong neural re-ranker before sparse or dense PRF can improve the retrieval effectiveness by 5-8%. This improvement in PRF effectiveness can be attributed directly to improving the precision of the feedback set. Second, we propose an enhanced expansion model, Latent Entity Expansion (LEE), which applies fine-grained word and entity-based relevance modelling incorporating localized features. Specifically, we find that by including both words and entities for expansion achieve a further 2-8% improvement in NDCG. Our analysis also demonstrated that LEE is largely robust to its parameters across datasets and performs well on entity-centric queries. And third, we include an 'adaptive' component in the retrieval process, which iteratively refines the re-ranking pool during scoring using the expansion model and avoids re-ranking additional documents. We find that this combination of techniques achieves the best NDCG, MAP and R@1000 results on the TREC Robust 2004 and CODEC document datasets, demonstrating a significant advancement in expansion effectiveness.

cs.IR↗

arXiv2023

GRM: Generative Relevance Modeling Using Relevance-Aware Sample Estimation for Document Retrieval

Recent studies show that Generative Relevance Feedback (GRF), using text generated by Large Language Models (LLMs), can enhance the effectiveness of query expansion. However, LLMs can generate irrelevant information that harms retrieval effectiveness. To address this, we propose Generative Relevance Modeling (GRM) that uses Relevance-Aware Sample Estimation (RASE) for more accurate weighting of expansion terms. Specifically, we identify similar real documents for each generated document and use a neural re-ranker to estimate their relevance. Experiments on three standard document ranking benchmarks show that GRM improves MAP by 6-9% and R@1k by 2-4%, surpassing previous methods.

cs.IR↗

arXiv2023

Generative and Pseudo-Relevant Feedback for Sparse, Dense and Learned Sparse Retrieval

Pseudo-relevance feedback (PRF) is a classical approach to address lexical mismatch by enriching the query using first-pass retrieval. Moreover, recent work on generative-relevance feedback (GRF) shows that query expansion models using text generated from large language models can improve sparse retrieval without depending on first-pass retrieval effectiveness. This work extends GRF to dense and learned sparse retrieval paradigms with experiments over six standard document ranking benchmarks. We find that GRF improves over comparable PRF techniques by around 10% on both precision and recall-oriented measures. Nonetheless, query analysis shows that GRF and PRF have contrasting benefits, with GRF providing external context not present in first-pass retrieval, whereas PRF grounds the query to the information contained within the target corpus. Thus, we propose combining generative and pseudo-relevance feedback ranking signals to achieve the benefits of both feedback classes, which significantly increases recall over PRF methods on 95% of experiments.

cs.IR↗