KARL: Knowledge Agents via Reinforcement Learning

• URL: http://arxiv.org/abs/2603.05218v1
• Date: Thu, 05 Mar 2026 14:30:25 GMT
• Title: KARL: Knowledge Agents via Reinforcement Learning
• Authors: Jonathan D. Chang, Andrew Drozdov, Shubham Toshniwal, Owen Oertell, Alexander Trott, Jacob Portes, Abhay Gupta, Pallavi Koppol, Ashutosh Baheti, Sean Kulinski, Ivan Zhou, Irene Dea, Krista Opsahl-Ong, Simon Favreau-Lessard, Sean Owen, Jose Javier Gonzalez Ortiz, Arnav Singhvi, Xabi Andrade, Cindy Wang, Kartik Sreenivasan, Sam Havens, Jialu Liu, Peyton DeNiro, Wen Sun, Michael Bendersky, Jonathan Frankle,
• Abstract summary: We present a system for training enterprise search agents via reinforcement learning.<n> KARLBench is a multi-capability evaluation suite spanning six distinct search regimes.<n>We show that models trained across heterogeneous search behaviors generalize substantially better than those optimized for any single benchmark.
• Score: 63.627906947205624
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: We present a system for training enterprise search agents via reinforcement learning that achieves state-of-the-art performance across a diverse suite of hard-to-verify agentic search tasks. Our work makes four core contributions. First, we introduce KARLBench, a multi-capability evaluation suite spanning six distinct search regimes, including constraint-driven entity search, cross-document report synthesis, tabular numerical reasoning, exhaustive entity retrieval, procedural reasoning over technical documentation, and fact aggregation over internal enterprise notes. Second, we show that models trained across heterogeneous search behaviors generalize substantially better than those optimized for any single benchmark. Third, we develop an agentic synthesis pipeline that employs long-horizon reasoning and tool use to generate diverse, grounded, and high-quality training data, with iterative bootstrapping from increasingly capable models. Fourth, we propose a new post-training paradigm based on iterative large-batch off-policy RL that is sample efficient, robust to train-inference engine discrepancies, and naturally extends to multi-task training with out-of-distribution generalization. Compared to Claude 4.6 and GPT 5.2, KARL is Pareto-optimal on KARLBench across cost-quality and latency-quality trade-offs, including tasks that were out-of-distribution during training. With sufficient test-time compute, it surpasses the strongest closed models. These results show that tailored synthetic data in combination with multi-task reinforcement learning enables cost-efficient and high-performing knowledge agents for grounded reasoning.

Related papers

• CoVe: Training Interactive Tool-Use Agents via Constraint-Guided Verification [17.56502992098113]
  We introduce textbfCoVe (textbfConstraint-textbfVerification), a post-training data synthesis framework designed for training interactive tool-use agents.<n>CoVe begins by defining explicit task constraints, which serve a dual role: they guide the generation of complex trajectories and act as deterministic verifiers for assessing trajectory quality.
  arXiv  Detail & Related papers  (2026-03-02T14:56:35Z)
• 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)
• Dual-Phase LLM Reasoning: Self-Evolved Mathematical Frameworks [48.105258051884384]
  This paper proposes a new two-stage training framework that enhances models' self-correction capabilities.<n>During the first stage, a multi-turn dialogue strategy guides the model to generate long chain-of-thought (CoT) data.<n>The second stage employs a difficulty-aware rejection sampling mechanism to dynamically optimize data distribution.
  arXiv  Detail & Related papers  (2026-01-09T08:19:11Z)
• Multimodal Reinforcement Learning with Agentic Verifier for AI Agents [131.46008226323423]
  Argos is a principled multimodal reward agent to train reasoning models for agentic tasks.<n>By leveraging our agentic verifier across both SFT data and RL training, our model achieves state-of-the-art results.
  arXiv  Detail & Related papers  (2025-12-03T04:42:47Z)
• Demystifying Reinforcement Learning in Agentic Reasoning [90.3737088727791]
  We conduct a comprehensive and systematic investigation to demystify reinforcement learning in agentic reasoning.<n>We highlight our key insights: (i) replacing stitched synthetic trajectories with real end-to-end tool-use trajectories yields a far stronger SFT.<n> Exploration-friendly techniques are crucial for agentic RL, such as clip higher, overlong reward shaping, and maintaining adequate policy entropy could improve the training efficiency.
  arXiv  Detail & Related papers  (2025-10-13T17:57:15Z)
• DSPO: Stable and Efficient Policy Optimization for Agentic Search and Reasoning [10.663331399616956]
  We introduce textbfDynamic-filter textbfSequence-level textbfPolicy textbfOptimization (DSPO), an improved RL algorithm for robust agent training.<n>We train our model purely through RL to interleave multi-turn search and reasoning, obviating the need for supervised demonstration data.
  arXiv  Detail & Related papers  (2025-10-10T10:53:25Z)
• Omni-Thinker: Scaling Multi-Task RL in LLMs with Hybrid Reward and Task Scheduling [66.0871543682453]
  We present Omni-Thinker, a unified reinforcement learning framework that scales large language models across diverse tasks.<n>Our scheduler orders tasks according to accuracy backward transfer (BWT), reducing forgetting and improving multi-task performance.
  arXiv  Detail & Related papers  (2025-07-20T01:50:16Z)
• LATTE: Learning to Think with Vision Specialists [110.43838069105998]
  We propose LATTE, a family of vision-language models that offload perception to state-of-the-art vision models.<n>By offloading perception to state-of-the-art vision models, our approach enables vision-language models to focus solely on reasoning over high-quality perceptual information.
  arXiv  Detail & Related papers  (2024-12-07T00:42:04Z)
• LamRA: Large Multimodal Model as Your Advanced Retrieval Assistant [63.28378110792787]
  We introduce LamRA, a versatile framework designed to empower Large Multimodal Models with sophisticated retrieval and reranking capabilities.<n>For retrieval, we adopt a two-stage training strategy comprising language-only pre-training and multimodal instruction tuning.<n>For reranking, we employ joint training for both pointwise and listwise reranking, offering two distinct ways to further boost the retrieval performance.
  arXiv  Detail & Related papers  (2024-12-02T17:10:16Z)
• When Parameter-efficient Tuning Meets General-purpose Vision-language Models [65.19127815275307]
  PETAL revolutionizes the training process by requiring only 0.5% of the total parameters, achieved through a unique mode approximation technique. Our experiments reveal that PETAL not only outperforms current state-of-the-art methods in most scenarios but also surpasses full fine-tuning models in effectiveness.
  arXiv  Detail & Related papers  (2023-12-16T17:13:08Z)
• Multi-Task Retrieval-Augmented Text Generation with Relevance Sampling [19.17759446168802]
  We study multi-task training of retrieval-augmented generation models for knowledge-intensive tasks. We filter training examples via a threshold of confidence on the relevance labels, whether a pair is answerable by the knowledge base or not.
  arXiv  Detail & Related papers  (2022-07-07T00:57:02Z)

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.