LongSeeker: Elastic Context Orchestration for Long-Horizon Search Agents

Overview

| Field | Content | |------|------| | Title | LongSeeker: Elastic Context Orchestration for Long-Horizon Search Agents | | Authors | Yijun Lu, Rui Ye, Yuwen Du, Jiajun Wang, Songhua Liu, Siheng Chen | | Published | 2026-05-06 | | Source | https://arxiv.org/abs/2605.05191 | | Type | Academic paper | | Section | Agentic Search |

One-Line Summary

This work introduces LongSeeker, a long-horizon search agent built on Context-ReAct, a paradigm for elastic, adaptive context orchestration that improves reliability and efficiency in agentic search.

Background and Motivation

Agentic search at scale has long faced challenges in efficiency, scalability, and user-intent understanding. Traditional pipelines separate retrieval, ranking, and generation, making it hard to meet modern demands for natural-language interaction, multi-hop reasoning, and real-time knowledge in the LLM era. LongSeeker targets this intersection.

The core problem: long-horizon search agents must manage a rapidly growing working context as they reason, call tools, and observe information. Naively accumulating all intermediate content can overwhelm the agent, increasing costs and the risk of errors.

Core Contributions

• Context-ReAct: a general agentic paradigm for elastic context orchestration integrating reasoning, context management, and tool use in a unified loop.
• Five atomic context operations — Skip, Compress, Rollback, Snippet, and Delete — enabling the agent to dynamically reshape its working context: preserving important evidence, summarizing resolved information, discarding unhelpful branches, and controlling context size.
• Theoretical guarantees: the Compress operator is proven expressively complete, while the other specialized operators provide efficiency and fidelity guarantees that reduce generation cost and hallucination risk.
• LongSeeker: a long-horizon search agent fine-tuned from Qwen3-30B-A3B on 10k synthesized trajectories.

Results

Across four representative search benchmarks, LongSeeker achieves:

• 61.5% on BrowseComp and 62.5% on BrowseComp-ZH
• Substantially outperforming Tongyi DeepResearch (43.2% and 46.7%) and AgentFold (36.2% and 47.3%)

These results highlight the potential of adaptive context management: agents can achieve more reliable and efficient long-horizon reasoning by actively shaping their working memory.

Key Insights for Search / Rec / Personalization

1. Architecture: cascaded retrieval + rerank + generation remains mainstream, but the agentic paradigm makes retrieval count and strategy themselves learnable. 2. Data: high-quality instruction data and session logs remain critical; synthetic data must guard against leakage and distribution shift. 3. Evaluation: the gap between offline metrics and online satisfaction is widening; LLM-as-judge needs cross-validation with human evaluation. 4. Product: latency, cost, explainability, and safety are hard constraints for industrial deployment.

Limitations and Future Work

Potential limitations include experiment scale constrained by GPU budget, benchmarks that differ from real user distributions, English-centric data with unknown cross-lingual generalization, and safety risks of agents on the open web. Future directions include more efficient test-time compute allocation, deeper integration with knowledge graphs and structured databases, and causal/fairness constraints for recommendation.

Original Abstract

> Long-horizon search agents must manage a rapidly growing working context as they reason, call tools, and observe information. Naively accumulating all intermediate content can overwhelm the agent, increasing costs and the risk of errors. We propose that effective context management should be adaptive: parts of the agent's trajectory are maintained at different levels of detail depending on their current relevance to the task. To operationalize this principle, we introduce Context-ReAct, a general agentic paradigm for elastic context orchestration that integrates reasoning, context management, and tool use in a unified loop. Context-ReAct provides five atomic operations: Skip, Compress, Rollback, Snippet and Delete, which allow the agent to dynamically reshape its working context, preserving important evidence, summarizing resolved information, discarding unhelpful branches, and controlling context size. We prove that the Compress operator is expressively complete, while the other specialized operators provide efficiency and fidelity guarantees that reduce generation cost and hallucination risk. Building on this paradigm, we develop LongSeeker, a long-horizon search agent fine-tuned from Qwen3-30B-A3B on 10k synthesized trajectories. Across four representative search benchmarks, LongSeeker achieves 61.5% on BrowseComp and 62.5% on BrowseComp-ZH, substantially outperforming Tongyi DeepResearch (43.2% and 46.7%) and AgentFold (36.2% and 47.3%). These results highlight the potential of adaptive context management, showing that agents can achieve more reliable and efficient long-horizon reasoning by actively shaping their working memory.