Laser: Governing Long-Horizon Agentic Search via Structured Protocol and Context Register

Source: arXiv:2512.20458 | Published: 2025-12-23 | Category: Agentic Search

Authors: Shuting Wang, Qiaolin Xia, Vich Wang, Herberttli, Bobsimons, Zhicheng Dou

Overview

Laser is a general framework for stabilizing and scaling agentic search — systems that interleave multi-step reasoning with external tool use. The work targets two chronic failure modes of LLM-driven search agents:

• Unstructured reasoning: reliance on free-form natural-language reasoning leads to unstable trajectories.
• Context overflow: raw intermediate traces accumulate in the context, degrading performance on complex multi-hop queries.

Key Components

1. Symbolic Action Protocol

Laser organizes agent behaviors into three action spaces:

• Planning – deciding what to do next
• Task-solving – executing retrieval and reasoning steps
• Retrospection – reviewing and correcting prior decisions

Each action has explicit semantics and a deterministic execution format, enabling:

• Structured, logical reasoning processes
• Reliable action parsing
• Interpretable and traceable intermediate decisions
• Fine-grained control over reasoning trajectories

2. Compact Context Register

In coordination with parsable actions, Laser maintains a compact register that stores only essential states of the reasoning process. This lets the agent reason over long horizons without uncontrolled context expansion.

Evaluation

Experiments were run on Qwen2.5/3-series models across challenging multi-hop QA datasets, comparing against existing agentic search baselines:

• Laser consistently outperforms baselines under both prompting-only and fine-tuning settings.

The authors conclude that Laser provides a principled and effective foundation for robust, scalable agentic search.

Original Abstract

> Recent advances in Large Language Models (LLMs) and Large Reasoning Models (LRMs) have enabled agentic search systems that interleave multi-step reasoning with external tool use. However, existing frameworks largely rely on unstructured natural-language reasoning and accumulate raw intermediate traces in the context, which often leads to unstable reasoning trajectories, context overflow, and degraded performance on complex multi-hop queries. In this study, we introduce Laser, a general framework for stabilizing and scaling agentic search. Laser defines a symbolic action protocol that organizes agent behaviors into three spaces: planning, task-solving, and retrospection. Each action is specified with explicit semantics and a deterministic execution format, enabling structured and logical reasoning processes and reliable action parsing. This design makes intermediate decisions interpretable and traceable, enhancing explicit retrospection and fine-grained control over reasoning trajectories. In coordination with parsable actions, Laser further maintains a compact context register that stores only essential states of the reasoning process, allowing the agent to reason over long horizons without uncontrolled context expansion. Experiments on Qwen2.5/3-series models across challenging multi-hop QA datasets show that Laser consistently outperforms existing agentic search baselines under both prompting-only and fine-tuning settings, demonstrating that Laser provides a principled and effective foundation for robust, scalable agentic search.

Context Within the Field

The paper sits at the intersection of agentic search and large-scale search systems, addressing how to redistribute the responsibilities of retrieval, ranking, generation, and tool calling in the LLM era. Related entries include:

• A Systematic Framework for Enterprise Knowledge Retrieval
• Synergizing RAG and Reasoning: A Systematic Review
• AceSearcher: Bootstrapping Reasoning and Search for LLMs via RL
• Agentic Information Retrieval

Note: Specific quantitative results should be verified against the original PDF; this article is based on the abstract and public metadata.