Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Table-as-Search: Formulate Long-Horizon Agentic Information Seeking as Table Completion

Tian Lan, Felix Henry, Bin Zhu, Qianghuai Jia, Junyang Ren, Qihang Pu, Haijun Li, Longyue Wang, Zhao Xu, Weihua Luo

TL;DR

TaS introduces Table-as-Search, a structured planning framework that reformulates long-horizon agentic InfoSeeking as Table Completion. By offloading search results to an external database and maintaining a structured table with rows for candidates and columns for constraints and information, TaS enables robust, scalable, and efficient management of complex, multi-hop information tasks. Across Deep, Wide, and DeepWide benchmarks, TaS consistently outperforms unstructured ReAct baselines and even commercial systems, while demonstrating improved robustness, higher precision-recall, and flexible integration of specialized sub-agents. The work highlights the architectural benefits of explicit state tracking and planning, offering a practical path toward industrial-scale, long-horizon information-seeking systems, with public code and datasets released for reproducibility.

Abstract

Current Information Seeking (InfoSeeking) agents struggle to maintain focus and coherence during long-horizon exploration, as tracking search states, including planning procedure and massive search results, within one plain-text context is inherently fragile. To address this, we introduce \textbf{Table-as-Search (TaS)}, a structured planning framework that reformulates the InfoSeeking task as a Table Completion task. TaS maps each query into a structured table schema maintained in an external database, where rows represent search candidates and columns denote constraints or required information. This table precisely manages the search states: filled cells strictly record the history and search results, while empty cells serve as an explicit search plan. Crucially, TaS unifies three distinct InfoSeeking tasks: Deep Search, Wide Search, and the challenging DeepWide Search. Extensive experiments demonstrate that TaS significantly outperforms numerous state-of-the-art baselines across three kinds of benchmarks, including multi-agent framework and commercial systems. Furthermore, our analysis validates the TaS's superior robustness in long-horizon InfoSeeking, alongside its efficiency, scalability and flexibility. Code and datasets are publicly released at https://github.com/AIDC-AI/Marco-Search-Agent.

Table-as-Search: Formulate Long-Horizon Agentic Information Seeking as Table Completion

TL;DR

TaS introduces Table-as-Search, a structured planning framework that reformulates long-horizon agentic InfoSeeking as Table Completion. By offloading search results to an external database and maintaining a structured table with rows for candidates and columns for constraints and information, TaS enables robust, scalable, and efficient management of complex, multi-hop information tasks. Across Deep, Wide, and DeepWide benchmarks, TaS consistently outperforms unstructured ReAct baselines and even commercial systems, while demonstrating improved robustness, higher precision-recall, and flexible integration of specialized sub-agents. The work highlights the architectural benefits of explicit state tracking and planning, offering a practical path toward industrial-scale, long-horizon information-seeking systems, with public code and datasets released for reproducibility.

Abstract

Current Information Seeking (InfoSeeking) agents struggle to maintain focus and coherence during long-horizon exploration, as tracking search states, including planning procedure and massive search results, within one plain-text context is inherently fragile. To address this, we introduce \textbf{Table-as-Search (TaS)}, a structured planning framework that reformulates the InfoSeeking task as a Table Completion task. TaS maps each query into a structured table schema maintained in an external database, where rows represent search candidates and columns denote constraints or required information. This table precisely manages the search states: filled cells strictly record the history and search results, while empty cells serve as an explicit search plan. Crucially, TaS unifies three distinct InfoSeeking tasks: Deep Search, Wide Search, and the challenging DeepWide Search. Extensive experiments demonstrate that TaS significantly outperforms numerous state-of-the-art baselines across three kinds of benchmarks, including multi-agent framework and commercial systems. Furthermore, our analysis validates the TaS's superior robustness in long-horizon InfoSeeking, alongside its efficiency, scalability and flexibility. Code and datasets are publicly released at https://github.com/AIDC-AI/Marco-Search-Agent.
Paper Structure (69 sections, 11 figures, 9 tables, 1 algorithm)

This paper contains 69 sections, 11 figures, 9 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Agentic Information Seeking.
  4. Agent Frameworks.
  5. Context Management.
  6. Task Formulation
  7. Problem Definition
  8. Table-as-Search (TaS) Framework
  9. Structured Schema Definition.
  10. Search as Table Completion.
  11. Unified View of InfoSeeking.
  12. Implementation of TaS Framework
  13. Table Initialization.
  14. Dynamic Orchestration.
  15. Answer Synthesis.
  16. ...and 54 more sections

Figures (11)

  • Figure 1: The overview of TaS Framework. Left: Unstructured planning (e.g., ReAct) is fragile and prone to massive context. Center: TaS reformulates InfoSeeking as Table Completion via row expansion and cell population. Right: TaS provides a unified representation for conducting Deep Search, Wide Search and DeepWide Search.
  • Figure 2: Robustness Analysis on BrowseComp-ZH (Top) and WideSearch (Bottom).
  • Figure 3: Search efficiency analysis of Gemini-2.5-Flash on Deep Search and Wide Search benchmarks.
  • Figure 4: Test-time Scaling Analysis on BrowseComp-ZH (Top, Gemini-2.5-Flash) and WideSearch (Bottom, Claude-Sonnet-4).
  • Figure 5: Case study between the ReAct and our proposed TaS Framework on the BrowseComp-ZH benchmark.
  • ...and 6 more figures