Polaris is presented, a system that trains an LLM to generate table descriptions directly from retrieval feedback, and demonstrates that retrieval benchmarks can be repurposed as supervision for training LLMs to generate retrieval-oriented metadata.
Abstract
Many table-centric NLP tasks such as NL2SQL first retrieve relevant tables from large collections using keyword search. Recent work uses LLMs to generate natural-language table descriptions to improve retrieval, but they are typically optimized for fluency rather than retrieval effectiveness. We present Polaris, a system that trains an LLM to generate table descriptions directly from retrieval feedback. Our key insight is that existing table retrieval benchmarks already contain the supervision needed for this task: given query-table relevance judgments, we generate multiple candidate descriptions for each table, rank them by their BM25 retrieval effectiveness, and use the resulting preference pairs to fine-tune the LLM with Direct Preference Optimization (DPO). Polaris further expands abbreviated table and column names before generation to reduce vocabulary mismatch. Extensive experiments show that Polaris outperforms the state-of-the-art AutoDDG solution, often by a significant margin. More broadly, our results demonstrate that retrieval benchmarks can be repurposed as supervision for training LLMs to generate retrieval-oriented metadata.
Evaluation on WikiSA and ExaRank shows that ranking-based few-shot prompting generally improves over zero-shot prompting and achieves competitive performance against random-shot prompting, indicating that retrieval-based demonstration selection is beneficial but not uniformly superior in all settings.
A. Laksito, Aali Alqarni, Mark Stevenson· International Conference on...· 0 citations
Tabular data represents a major source of structured knowledge for open-domain question answering (QA) and enterprise data lakes, yet effective table retrieval remains challenging due to the structure–semantics gap imposed by tabular layouts. Heuristic table linearization often leads to semantic loss, particularly for implicit queries, while recent supervised retrieval models (e.g., Birdie, Contr) rely heavily on large-scale query logs and labeled QA pairs, limiting their applicability in Day-0 cold-start scenarios. We propose GeCo-TR (Generative Schema and Contrastive Table Retrieval), a zero-shot table retrieval framework that eliminates the need for supervised QA data by shifting from direct query-to-table learning to modeling the intrinsic structural semantics of the table corpus. GeCo-TR introduces UHMI, a unified hybrid representation that integrates table structure with linked knowledge graph entities, and employs a hybrid neural–symbolic retrieval mechanism that dynamically combines dense semantic retrieval, symbolic graph traversal, and sparse lexical matching. This design enables robust semantic generalization while enforcing explicit structural constraints, resulting in high-precision and high-recall retrieval for implicit queries in a zero-shot setting. Extensive experiments on public benchmarks demonstrate the effectiveness of GeCo-TR; notably, on Open-WikiTable, it achieves 97.5% Recall@5 in the zero-shot setting, ranking second among all evaluated methods despite requiring no query-level supervision.
Zhou He, Zhifei Pang, Xiu Tang et al.· Annual International ACM SIG...· 0 citations
This work introduces AnchorQE, a training-free method that separately encodes the original query and its expansion before interpolating them, and shows that AnchorQE improves retrieval effectiveness by up to 12.89% when compared to widely-used expansion-only or text-level concatenation baselines across TREC-DL, LoTTE, and BEIR.
Lexical retrieval (BM25) captures exact keyword matches and weights terms by corpus-wide significance, but it is blind to the semantic vocabulary gap: when a relevant document phrases an answer differently from the query, BM25 never retrieves it, and no amount of downstream reranking or fusion can recover a document that was never in the candidate set. We present Cross-Encoder Query Expansion (CE-QE), which reads the per-token relevance attributions of a cross-encoder applied to top semantic search results, selects the terms the cross-encoder treats as decisive, and appends them to the BM25 query. Unlike classical pseudo-relevance feedback, which reuses BM25's own (possibly wrong) top results, CE-QE seeds expansion from the semantic retriever's results, avoiding self-reinforcing query drift. Unlike recent generative query expansion (HyDE, Query2doc), which prompts a large language model to hallucinate text from its parametric knowledge, every CE-QE expansion term is copied verbatim from a retrieved passage, so it cannot introduce vocabulary the corpus does not contain, and its only added cost is attribution extraction on a cross-encoder a hybrid pipeline already runs for reranking. On seven BEIR datasets, CE-QE improves lexical recall substantially where query and answer vocabulary diverge (e.g., NQ Recall@100 from 0.32 to 0.47), and its score-fusion variant (SESF) beats cross-encoder score fusion by 2.5% on Recall@100 and beats SPLADEv2 and ColBERTv2 by 5.3% and 4.6% on nDCG@10, while leaving the underlying BM25 index completely unmodified.
Adam Kahirov, Umesh Deshpande, S. Sundararaman· 0 citations
This paper builds directly on our prior work presented at ICES 2025, which introduced a hybrid retrieval system combining sparse lexical retrieval (BM25), dense vector embeddings, and Reciprocal Rank Fusion to streamline NASA research search and retrieval. In that work, experimental results revealed a critical limitation: short-form queries, typical of user search behavior, consistently underperformed longer, semantically-rich queries across all retrieval strategies, with sparse retrieval exhibiting the largest degradation. This performance gap was attributed to a fundamental mismatch between the brevity of user queries and the length, jargon density, and conceptual structure of technical abstracts within the corpus. To address this limitation, this work proposes and evaluates the integration of a Hypothetical Document Embeddings (HyDE) layer into the retrieval pipeline. Rather than embedding the raw user query, the system first generates a hypothetical abstract that reflects the theoretical content, terminology, and structure of a relevant technical paper's abstract, corresponding to the intent of the user’s query. This generated abstract is then used as the retrieval query for both sparse and dense search methods. By increasing semantic density and domain-specific language, the hypothetical document theoretically improves alignment with indexed abstracts in both term-frequency and embedding space. We integrate this HyDE-based approach into the existing modular hybrid retrieval architecture and evaluate its impact on retrieval effectiveness across varying query lengths and retrieval strategies. Contrary to expectations, the results demonstrate that hypothetical document generation does not improve ranking performance for short-form or long-form queries.
Lily Hanslik, Ian Ortega, Hunter Sandidge et al.· 55th International Conferenc...· 0 citations
Results demonstrate a 157% improvement in retrieval quality over a Naive-RAG baseline, with only 50 ms additional la tency, while Qwen2-VL-2B-Instruct achieved results comparable to cloud-based models in BERTScore, validate that open-source optimized SLMs, paired with advanced retrieval strategies, can provide competitive performance for document understanding without relying on cloud-based models.