---
title: 'Table-as-Search (TaS): Structured Table Retrieval'
url: https://www.emergentmind.com/topics/table-as-search-tas
type: topic
---

# Table-as-Search (TaS): Structured Table Retrieval

Searching arXiv for recent papers on “Table-as-Search” and closely related table retrieval formulations.
Table-as-Search (TaS) denotes a family of formulations in which tables, or table-derived structures, are treated as the primary search objects rather than as incidental context for document retrieval or SQL execution. In current usage, the term covers at least two closely related lines of work: retrieval over corpora or data lakes where the output is one or more relevant tables, and a structured planning framework that reformulates long-horizon information seeking as table completion [2511.13418][2605.18199][2602.06724]. Across both lines, the central design move is to preserve structure—fields, rows, columns, entities, joins, alignments, or pending cells—so that search can optimize semantic relevance together with relational organization.

## 1. Conceptual scope and research lineage

Early table search work treated a table as a structured object with heterogeneous evidence sources rather than as a flat text fragment. In scientific table search, MaitreD decomposed each table into document-level fields, table-level fields, and cell-level fields, then ranked by a probabilistic framework of the form $p(t \mid q) \propto p(t)\,p(q \mid t)$, where the prior $p(t)$ favored numeric-content-rich tables [1707.03423]. The same paper also introduced TableArXiv, built from the arXiv January 2014 snapshot in Physics and related categories, with 341,573 extracted tables and 105 expert-created topics [1707.03423].

Neural retrieval work shifted the emphasis from explicit field weighting to learned representation. BERT-based ad hoc table retrieval adapted structured tables to a token-limited encoder by selecting rows, columns, or cells and concatenating them with table context fields; on WikiTables, “Hybrid-BERT-Row-Max” reached MAP \(=0.6311\), MRR \(=0.6673\), and NDCG@10 \(=0.6519\), exceeding STR and simpler BERT baselines [2005.09207]. StruBERT then fused textual fields and tabular structure by combining row-based and column-based linearizations with horizontal and vertical self-attention, and improved content-based retrieval on Query-by-Example from TaBERT’s NDCG@5 \(=0.5877\) to \(0.6345\) [2203.14278].

More recent work has made the field heterogeneity of tables explicit again, but within neural and hybrid frameworks. THYME argued that titles benefit from coarse-grained semantic matching, whereas headers and cells often require fine-grained lexical matching, and implemented this with a dense semantic score plus a field-aware sparse lexical score controlled by a Mixture-of-Field-Experts gate [2503.02251]. PIPER, EasyTUS, TACTUS, GenTUS, and related systems extended TaS from web-style table retrieval to large, metadata-poor data lakes in which the dominant tasks are dataset search, union search, join-aware composition, and multi-table retrieval [2605.18199][2511.02674][2603.17298][2607.00833][2511.13418].

Taken together, these works indicate that TaS is not a single retrieval model but a design paradigm whose defining property is that table structure remains operational throughout ranking, candidate generation, or planning.

## 2. Query modalities and formal problem statements

The TaS literature spans several distinct query modalities and output spaces.

| Task family | Query object | Representative formulations |
|---|---|---|
| Keyword or natural-language table retrieval | Text query | Scientific table search [1707.03423], PIPER dataset search [2605.18199] |
| Related-table or query-by-example retrieval | Table query | OGDC relatedness tasks [2308.13560], StruBERT content-based retrieval [2203.14278] |
| Table union search | Query table | EasyTUS [2511.02674], TACTUS [2603.17298], GenTUS [2607.00833] |
| Multi-table retrieval | Natural-language query | Iterative join-aware retrieval for NL2SQL [2511.13418] |
| Novel table search | Query table plus unionable candidates | ANTs and NTS variants [2603.07235] |
| Long-horizon information seeking as table completion | Natural-language query | TaS planner over external table state [2602.06724] |

In keyword-style settings, the query is free text and the output is a ranked list of tables whose textual, structural, or content-derived features are relevant. MaitreD modeled this as structured probabilistic retrieval over eight fields, while PIPER defined the TaS task as ranking tables whose content and structure make them relevant to a free-form natural-language query for downstream analysis, integration, or exploration [1707.03423][2605.18199].

In table-to-table settings, the query itself is a table. OGDC formalized three notions of relatedness—same dataset, same organization, and weighted tag overlap—and provided both related-table search and keyword-search benchmarks over roughly \(195{,}448\) distinct tables collected from seven CKAN portals [2308.13560]. In table union search, the objective is to find top-\(k\) tables whose union with a query table produces a semantically consistent extension of rows; this is the task addressed by EasyTUS, TACTUS, and GenTUS [2511.02674][2603.17298][2607.00833].

Composition-oriented TaS generalizes beyond returning a single table. In “Exploring Multi-Table Retrieval Through Iterative Search,” the input is a natural-language query \(Q\) and a datalake of tables \(\mathcal{T}=\{T_1,\ldots,T_N\}\), and the output is a small set \(S \subseteq \mathcal{T}\), \(|S| \le K\), whose joint content can answer \(Q\). The paper defines a joint score that balances semantic relevance, coverage of query sub-concepts, and structural coherence via joinability [2511.13418]. Novel Table Search introduces a second-stage ranking problem over unionable tables, seeking tables that contribute new information relative to the query table rather than near-duplicates of it [2603.07235].

A separate but increasingly cited use of the term redefines search itself as a table-completion process. In that framework, a query is mapped to a schema \(S=\langle K,C,I\rangle\), rows correspond to candidate entities, columns correspond to constraints or information to be collected, filled cells record evidence, and empty cells encode an explicit search plan [2602.06724]. This suggests that “Table-as-Search” now names both a retrieval paradigm and a state-management paradigm.

## 3. Representation and ranking architectures

A recurrent issue in TaS is how much of a table to encode and at what granularity. The earliest probabilistic systems handled this by field decomposition and smoothing. MaitreD assigned different emphasis to article title, abstract, caption, referring sentences, footnotes, column headers, row headers, and cell values, and additionally used entity extraction and quantity-type expansion via QUDT units to bridge query–table mismatch [1707.03423].

Transformer-based retrievers initially confronted the token-budget problem directly. “Table Search Using a Deep Contextualized Language Model” proposed a content selector that ranks rows, columns, or cells by salience relative to the query using fastText-based similarity, after which only the top-ranked items are serialized into BERT together with caption, section title, page title, and header row [2005.09207]. The reported ablations found row-based selection superior to column-based or cell-based alternatives, and max-salience strongest for rows [2005.09207].

Structure-aware encoders then made row and column organization part of the representation itself. StruBERT linearized each table into column-based sequences and row-based sequences, pooled BERT token states into cell embeddings, applied vertical and horizontal self-attention, and compared tables through both coarse-grained \([CLS]\)-style signals and fine-grained row/column interactions computed with a lightweight miniBERT layer [2203.14278]. THYME followed a different route: a shared Transformer produces both a dense semantic vector and sparse lexical logits, title pooling uses max aggregation, header and cell pooling are field-specific, and a gating network dynamically weights title, header, and cell evidence before the final hybrid score \(s(q,t)=s_{sem}+s_{lex}\) at inference [2503.02251].

Data-lake systems increasingly avoid publisher metadata altogether. PIPER profiles each table statistically, generating per-column features such as inferred datatype, number of distinct values, fraction of missing entries, coverage span, and type-specific statistics; an LLM then turns the serialized profile into multiple pseudoqueries, which are embedded and indexed for dense retrieval [2605.18199]. TUNES takes an entity-oriented route: it identifies primary keys and attributes, instantiates a Neo4j property graph, combines BM25 over entity documents with semantic search using bge-m3 embeddings, and augments the retrieved set with entities returned by LLM-generated Cypher queries [2508.17028].

A plausible implication is that TaS representation learning has diverged into three stable families: field-aware text ranking, holistic table embedding, and structure-preserving entity or graph indexing. The literature does not treat these families as mutually exclusive; several systems combine them.

## 4. Composition-oriented search in data lakes

Data-lake TaS places particular emphasis on compositionality: whether retrieved tables can be unioned, joined, or otherwise combined into a coherent answer.

In Table Union Search, EasyTUS implements a three-step pipeline of table serialization, table representation, and vector search. It uses a zero-shot, value-only serialization strategy with random row sampling, mean-pooled LLM embeddings, and HNSW-based approximate nearest-neighbor retrieval by cosine similarity [2511.02674]. TACTUS argues that prior TUS methods are overly column-centric and instead learns one embedding per table to capture holistic unionability, using positive table pair construction, latent-positive exclusion, hard-negative mining, attentive table encoding, adaptive candidate retrieval, and dual-evidence reranking that combines table-level and column-level signals [2603.17298].

GenTUS departs from the standard encode-search-rerank architecture altogether. It assigns each table a unionability-aware semantic identifier by residual quantization of learned table embeddings, trains an encoder–decoder model to generate the identifiers of unionable tables directly from a serialized query table, and enforces validity at inference through constrained decoding over a prefix trie of known identifiers [2607.00833]. Novel Table Search adds yet another objective: among already unionable tables, prefer those that are syntactically novel. ANTs approximates novelty through attribute-level scores that combine low syntactic similarity with semantic similarity, and ranks candidate tables by \(\mathrm{TableNovelty}(Q,T)\) [2603.07235].

Join-oriented composition introduces a different constraint. The iterative multi-table retrieval framework for open-domain question answering over data lakes defines a joint objective over relevance, coverage, and joinability and explicitly treats retrieval as an iterative search process. Its concrete instantiation, Greedy Join-Aware Retrieval, uses pre-computed coarse relevance \(r_i\), fine-grained relevance matrix \(F_{ji}\), and join compatibility \(\omega_{il}\), with a utility that sums coarse, coverage, and join gains using weights \(\lambda_{coarse}\), \(\lambda_{cov}\), and \(\lambda_{join}\) [2511.13418].

These systems share a common departure from classical document retrieval: admissible results are not merely relevant in isolation. They must also support a downstream table operation—union, join, augmentation, or novelty-aware diversification.

## 5. Benchmarks and empirical findings

Benchmark design has been central to TaS because evaluation depends strongly on the task definition. TableArXiv targeted scientific keyword search with graded relevance labels over 105 topics [1707.03423]. OGDC scaled evaluation to approximately \(195{,}448\) distinct open-government tables and defined automatic labels for related-table search and keyword search via metadata such as dataset, organization, and tags [2308.13560]. Query-by-Example, WikiTables, PMC, NQ-TABLES, OTT-QA, FetaQA, NTCIR-15, SPIDER, BIRD, FIBEN, BEAVER-DW, BEAVER-NW, TUSBench, WikiUnion, LakeBench, GAIA, BrowseComp-ZH, and WideSearch subsequently extended TaS evaluation across web tables, data lakes, and long-horizon search settings [2203.14278][2503.02251][2605.18199][2511.13418][2511.02674][2607.00833][2602.06724].

Reported gains are substantial but task-specific. In scientific table search, full MaitreD improved over BM25F from NDCG@20 \(=0.2671\) to \(0.3666\), ERR@20 \(=0.1070\) to \(0.1495\), and MAP@100 \(=0.2025\) to \(0.2784\) [1707.03423]. On OGDC related-table search over the Canada site, Siamese reached NDCG@20 \(=85.1\%\) for same-dataset relatedness, \(94.8\%\) for same-organization relatedness, and \(27.8\%\) for tag-overlap relatedness, all above BM25 [2308.13560]. THYME achieved NQ-TABLES NDCG@5 \(=65.7\), NDCG@10 \(=68.1\), and R@1 \(=48.6\), and OTT-QA NDCG@5 \(=78.2\), NDCG@10 \(=79.6\), and R@1 \(=66.7\), significantly outperforming BIBERT+SPLADE [2503.02251].

For metadata-poor dataset search, PIPER obtained FetaQA R@10 \(=0.784\) versus \(0.586\) for pT+QGpT, while on OTT-QA the metadata-strong setting favored baselines, with PIPER at \(0.729\) versus \(0.915\) [2605.18199]. On NTCIR-15 complex natural-language queries, PIPER reported MAP \(=0.560\), P@10 \(=0.480\), R@10 \(=0.647\), and nDCG@10 \(=0.676\), all above Dense-BGE [2605.18199].

In union search, EasyTUS reported average MAP improvements of up to \(34.3\%\), data-preparation speedups up to \(79.2\times\), and query-processing speedups up to \(7.7\times\), while retaining \(>95\%\) of its accuracy in metadata-absent variants [2511.02674]. TACTUS achieved MAP@k \(\ge 0.9860\) on SANTOS Small, \(0.9936\) on TUS Small, \(0.9844\) on TUS Large, and \(0.6537\) on WikiUnion, while reducing candidate set size and improving purity [2603.17298]. GenTUS reported the best overall retrieval quality across seven public TUS benchmarks, with average rank \(\bar r=1.05\) compared to \(2.57\) for TACTUS, as well as \(10\text{–}200\times\) faster online retrieval and \(10\text{–}60\times\) smaller storage [2607.00833].

In multi-table retrieval for NL2SQL, the iterative join-aware method was competitive with a mixed-integer-programming baseline while being much faster: on SPIDER at \(K=2\), greedy obtained \(85.5\%\) Recall and \(68.0\%\) Complete Recall in about \(20\) s versus \(325\) s; on BIRD at \(K=3\), \(84.7\%/70.5\%\) versus \(85.6\%/71.9\%\) in \(101\) s versus \(13{,}679\) s; and on BEAVER-NW at \(K=5\), \(49.6\%/10.5\%\) versus \(49.2\%/0.0\%\) in \(30\) s versus \(7{,}297\) s [2511.13418].

The main empirical lesson is not that one architecture dominates universally, but that gains are largest when the retrieval objective aligns tightly with the downstream table operation and with the actual evidence available in the corpus.

## 6. Table completion as long-horizon search state

A distinct research line uses the phrase “Table-as-Search” for long-horizon agentic information seeking rather than corpus retrieval. In that formulation, a deterministic schema-construction function maps a query \(q\) to a schema \(S=\langle K,C,I\rangle\), where \(K\) is a set of candidate entities, \(C\) is a set of constraint columns, and \(I\) is a set of information columns; at time \(t\), the system maintains a table \(T_t \in (V \cup \{\emptyset,\mathrm{N/A}\})^{m \times (p+q)}\) whose empty cells denote pending search steps [2602.06724].

The framework unifies three paradigms by schema choice: Deep Search when \(|C|>0, |I|=0\), Wide Search when \(|C|=0, |I|>0\), and DeepWide Search when both are nonempty [2602.06724]. The planner alternates between expanding rows and populating empty cells, while the external table persists search state that would otherwise be compressed into a fragile plain-text context window [2602.06724].

The reported results indicate that this state externalization matters. On BrowseComp-ZH, TaS-MA with Gemini-2.5-Flash improved from \(38.4\%\) to \(52.4\%\), and with GPT-5 Medium-Think from \(71.8\%\) to \(77.7\%\), relative to ReAct-MA [2602.06724]. On DeepWide Search over 20 e-commerce BD queries, TaS-MA with Claude-S4 reached Col-F1 \(=55.9\%\) and Item-P \(=63.5\%\), above Gemini DeepResearch and ReAct-MA [2602.06724]. The authors also reported that the performance gap versus ReAct-MA widens as query complexity increases, and that Planner quality is the primary bottleneck [2602.06724].

This usage broadens TaS from “search over tables” to “search through a table-shaped state space.” It remains structurally consistent with retrieval-oriented TaS because the table is again the mechanism that organizes evidence and constrains admissible next actions.

## 7. Limitations, ambiguities, and open problems

Several limitations recur across the literature. Metadata sparsity remains a major problem for data-lake search, motivating systems such as PIPER and EasyTUS that either replace metadata with content-derived profiles or ignore metadata entirely [2605.18199][2511.02674]. Token-budget constraints remain a structural issue for Transformer encoders, and early BERT-based work still relied on fixed salience heuristics that cannot dynamically choose the optimal granularity of rows, columns, or cells per query [2005.09207]. In TUNES, graph-search errors accounted for approximately \(50\%\) of search errors, while entity identification failed on highly irregular table layouts in \(4\%\) of analyzed cases [2508.17028].

LLM-dependent TaS systems introduce additional concerns: pseudoquery generation and query optimization incur quality and cost dependencies, and synthetic query generation may hallucinate [2605.18199]. EasyTUS reported that on the purely numeric-heavy ecb_union dataset, D3L slightly outperformed its LLM-based approach, which the paper identifies as a known limitation of pure-LLM methods [2511.02674]. GenTUS reduces index size and latency, but its own discussion notes that a generative retriever depends on a well-structured identifier space and that beam search may struggle if the codebook does not align with semantics [2607.00833].

Complexity-theoretic limits also appear repeatedly. The one-shot mixed-integer formulation for multi-table retrieval has \(O(2^N)\) worst-case complexity, motivating greedy iterative approximations [2511.13418]. Novel Table Search similarly proves the associated optimization problem NP-hard and therefore turns to penalization-based approximation through ANTs [2603.07235]. In the long-horizon planning setting, planner degradation rather than sub-agent degradation produced the largest drops, indicating that search control itself can dominate tool quality [2602.06724].

A persistent ambiguity is terminological. “Table-as-Search” now refers both to retrieval of tables as indexed objects and to the reformulation of search as table completion. This suggests a broader unifying interpretation: TaS is any search system in which table structure is not merely input formatting but the principal computational substrate for ranking, composition, or control.

Source: https://www.emergentmind.com/topics/table-as-search-tas