CRISP: Critical Step Perception for Training Efficient Deep Search Agents

• URL: http://arxiv.org/abs/2608.01867v2
• Date: Tue, 04 Aug 2026 02:03:12 GMT
• Title: CRISP: Critical Step Perception for Training Efficient Deep Search Agents
• Abstract summary: We propose CRISP, a framework for training efficient deep search agents through critical step perception.<n>Unlike prior efficiency methods that uniformly penalize tool use, CRISP distinguishes interactions that gather necessary evidence from redundant ones.<n>Experiments show that CRISP maintains competitive final-answer accuracy while reducing average interaction turns by 15.1% and 33.2%, respectively.
• Score: 19.74419958232817
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large language models (LLMs) are increasingly extended into deep search agents that solve complex questions through multi-step interaction with external search and browsing tools. However, existing agents often incur substantial computational and interaction costs, generating lengthy trajectories that contain redundant queries, inefficient exploration, and irrelevant observations. Existing efficiency-oriented methods usually encourage agents to use tools less frequently, but treating all tool interactions uniformly may also suppress steps that gather necessary evidence. In this paper, we propose CRISP, a framework for training efficient deep search agents through critical step perception. Unlike prior efficiency methods that uniformly penalize tool use, CRISP distinguishes interactions that gather necessary evidence from redundant ones and shapes the training reward to preserve the former while pruning the latter, improving efficiency without sacrificing the evidence needed for correct answers. Specifically, CRISP first constructs critical-step labels with Backward Evidence Induction: starting from the final answer, a strong model traverses a completed search trajectory backward and judges whether each tool-interaction step provides or preserves evidence for the final answer. We then distill these step-wise judgments into a smaller critical-step recognizer, enabling full-trajectory analysis in a single pass. During policy optimization, an efficiency-aware reward is applied only to successful rollouts. Experiments on BrowseComp and HLE-Verified show that CRISP maintains competitive final-answer accuracy while reducing average interaction turns by 15.1% and 33.2%, respectively, demonstrating substantial improvements in interaction efficiency.

Related papers

• HyperEyes: Dual-Grained Efficiency-Aware Reinforcement Learning for Parallel Multimodal Search Agents [6.536862833942476]
  We present HyperEyes, a parallel multimodal search agent that fuses visual grounding and retrieval into a single atomic action.<n>Across six benchmarks, HyperEyes-30B surpasses the strongest comparable open-source agent by 9.9% in accuracy with 5.3x fewer tool-call rounds on average.
  arXiv  Detail & Related papers  (2026-05-08T03:16:08Z)
• CoSearch: Joint Training of Reasoning and Document Ranking via Reinforcement Learning for Agentic Search [51.911048955965136]
  CoSearch is a framework that jointly trains a multi-step reasoning agent and a generative document ranking model.<n>Our results show that joint training of the reasoning agent and retrieval system is both feasible and strongly performant.
  arXiv  Detail & Related papers  (2026-04-19T17:48:17Z)
• AutoSearch: Adaptive Search Depth for Efficient Agentic RAG via Reinforcement Learning [52.305422887002656]
  We propose a reinforcement learning framework that evaluates each search step via self-generated intermediate answers.<n>By a self-answering mechanism, AutoSearch identifies the minimal sufficient search depth and promotes efficient search.<n>Experiments show that AutoSearch achieves a superior accuracy-efficiency trade-off, alleviating over-searching while preserving search quality.
  arXiv  Detail & Related papers  (2026-04-19T09:05:48Z)
• PRAISE: Prefix-Based Rollout Reuse in Agentic Search Training [28.912613644535668]
  We present PRAISE, a framework for improving data efficiency and credit assignment in agentic search training.<n>Our method uses a single shared model for both search policy learning and prefix answer evaluation.<n>Experiments on multi-hop QA benchmarks show that PRAISE consistently improves performance over strong baselines.
  arXiv  Detail & Related papers  (2026-04-04T10:23:46Z)
• Learning to Retrieve from Agent Trajectories [72.8923565916533]
  We argue that retrieval models for agentic search should be trained directly from agent interaction data.<n>We introduce learning to retrieve from agent trajectories as a new training paradigm, where supervision is derived from multi-step agent interactions.<n>Our results highlight agent trajectories as a practical and scalable supervision source, pointing to a promising direction for retrieval in the era of agentic search.
  arXiv  Detail & Related papers  (2026-03-30T17:59:02Z)
• SRR-Judge: Step-Level Rating and Refinement for Enhancing Search-Integrated Reasoning in Search Agents [30.92763154920672]
  We introduce SRR-Judge, a framework for reliable step-level assessment of reasoning and search actions.<n>SRR-Judge provides fine-grained guidance for search-integrated reasoning and enables efficient post-training annotation.<n> Empirically, SRR-Judge delivers more reliable step-level evaluations than much larger models such as DeepSeek-V3.1.
  arXiv  Detail & Related papers  (2026-02-08T02:07:41Z)
• Search-R2: Enhancing Search-Integrated Reasoning via Actor-Refiner Collaboration [49.9937230730202]
  We propose Search-R2, a novel Actor-Refiner collaboration framework that enhances reasoning through targeted intervention.<n>Our approach decomposes the generation process into an Actor, which produces initial reasoning trajectories.<n>We show that Search-R2 consistently outperforms strong RAG and RL-based baselines across model scales.
  arXiv  Detail & Related papers  (2026-02-03T15:32:09Z)
• Thinking Forward and Backward: Multi-Objective Reinforcement Learning for Retrieval-Augmented Reasoning [137.33138614095435]
  Retrieval-augmented generation (RAG) has proven to be effective in mitigating hallucinations in large language models.<n>Recent efforts have incorporated search-based interactions into RAG, enabling iterative reasoning with real-time retrieval.<n>We propose Bi-RAR, a novel retrieval-augmented reasoning framework that evaluates each intermediate step jointly in both forward and backward directions.
  arXiv  Detail & Related papers  (2025-11-12T08:29:39Z)
• WebLeaper: Empowering Efficiency and Efficacy in WebAgent via Enabling Info-Rich Seeking [60.35109192765302]
  Information seeking is a core capability that enables autonomous reasoning and decision-making.<n>We propose WebLeaper, a framework for constructing high-coverage IS tasks and generating efficient solution trajectories.<n>Our method consistently achieves improvements in both effectiveness and efficiency over strong baselines.
  arXiv  Detail & Related papers  (2025-10-28T17:51:42Z)
• Beyond Outcome Reward: Decoupling Search and Answering Improves LLM Agents [19.31471304268234]
  We introduce DeSA (Decoupling Search-and-Answering), a simple two-stage training framework that explicitly separates search optimization from answer generation.<n>Across seven QA benchmarks, DeSA-trained agents consistently improve search behaviors, delivering substantially higher search recall and answer accuracy than outcome-only baselines.
  arXiv  Detail & Related papers  (2025-10-06T11:09:45Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.