mlsanityData to 14 Oct 2026

Scaling Agents via Continual Pre-training

arXiv 2509.13310 ↗First posted 16 Sep 2025cs.CLcited by 40

On the map: LLM agents and tool use → General LLM agent frameworks and benchmarks

Abstract (arXiv)

Large language models (LLMs) have evolved into agentic systems capable of autonomous tool use and multi-step reasoning for complex problem-solving. However, post-training approaches building upon general-purpose foundation models consistently underperform in agentic tasks, particularly in open-source implementations. We identify the root cause: the absence of robust agentic foundation models forces models during post-training to simultaneously learn diverse agentic behaviors while aligning them to expert demonstrations, thereby creating fundamental optimization tensions. To this end, we are the first to propose incorporating Agentic Continual Pre-training (Agentic CPT) into the deep research agents training pipeline to build powerful agentic foundational models. Based on this approach, we develop a deep research agent model named AgentFounder. We evaluate our AgentFounder-30B on 10 benchmarks and achieve state-of-the-art performance while retains strong tool-use ability, notably 39.9% on BrowseComp-en, 43.3% on BrowseComp-zh, and 31.5% Pass@1 on HLE.

Key facts. Cited by 40 papers in the map (data through 14 Oct 2026). Placed in General LLM agent frameworks and benchmarks, one of 16 directions in LLM agents and tool use. Builds on 38 of its 61 references that are in the map. First posted on arXiv 16 Sep 2025.

Builds on

38 of 61 references.

Closest ancestors · 2

Same direction, newest first

  1. 2025GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models8 Aug
  2. 2025Chain-of-Agents: End-to-End Agent Foundation Models via Multi-Agent Distillation and Agentic RL6 Aug

Other references · 36

  1. 2022Training language models to follow instructions with human feedbackcited by 15,971
  2. 2025Qwen3 Technical Reportcited by 8,820
  3. 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,695
  4. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,330
  5. 2024Qwen2.5 Technical Reportcited by 4,942
  6. 2023Toolformer: Language Models Can Teach Themselves to Use Toolscited by 4,182
  7. 2022Scaling Instruction-Finetuned Language Modelscited by 3,024
  8. 2023GAIA: a benchmark for General AI Assistantscited by 1,128
  9. 2025BrowseComp: A Simple Yet Challenging Benchmark for Browsing Agentscited by 647
  10. 2025Humanity's Last Examcited by 491
  11. 2025Search-o1: Agentic Search-Enhanced Large Reasoning Modelscited by 433
  12. 2025R1-Searcher: Incentivizing the Search Capability in LLMs via Reinforcement Learningcited by 336
24 more
  1. 2025WebThinker: Empowering Large Reasoning Models with Deep Research Capabilitycited by 335
  2. 2023Tool Learning with Foundation Modelscited by 302
  3. 2025DeepResearch Bench: A Comprehensive Benchmark for Deep Research Agentscited by 248
  4. 2024Continual Learning of Large Language Models: A Comprehensive Surveycited by 234
  5. 2025WebSailor: Navigating Super-human Reasoning for Web Agentcited by 229
  6. 2023Continual Pre-training of Language Modelscited by 198
  7. 2025WebDancer: Towards Autonomous Information Seeking Agencycited by 188
  8. 2024Fact, Fetch, and Reason: A Unified Evaluation of Retrieval-Augmented Generationcited by 162
  9. 2025Beyond Ten Turns: Unlocking Long-Horizon Agentic Search with Large-Scale Asynchronous RLcited by 150
  10. 2025BrowseComp-ZH: Benchmarking Web Browsing Ability of Large Language Models in Chinesecited by 145
  11. 2023Continual Pre-Training of Large Language Models: How to (re)warm your model?cited by 137
  12. 2025WebShaper: Agentically Data Synthesizing via Information-Seeking Formalizationcited by 121
  13. 2025WebWatcher: Breaking New Frontier of Vision-Language Deep Research Agentcited by 119
  14. 2021Lifelong Pretraining: Continually Adapting Language Models to Emerging Corporacited by 115
  15. 2025WebExplorer: Explore and Evolve for Training Long-Horizon Web Agentscited by 92
  16. 2025SealQA: Raising the Bar for Reasoning in Search-Augmented Language Modelscited by 75
  17. 2024Reuse, Don't Retrain: A Recipe for Continued Pretraining of Language Modelscited by 72
  18. 2025ACEBench: Who Wins the Match Point in Tool Usage?cited by 71
  19. 2024Investigating Continual Pretraining in Large Language Models: Insights and Implicationscited by 57
  20. 2025DeepDive: Advancing Deep Search Agents with Knowledge Graphs and Multi-Turn RLcited by 55
  21. 2025EvolveSearch: An Iterative Self-Evolving Search Agentcited by 43
  22. 2025Cognitive Kernel-Pro: A Framework for Deep Research Agents and Agent Foundation Models Trainingcited by 42
  23. 2025MaskSearch: A Universal Pre-Training Framework to Enhance Agentic Search Capabilitycited by 22
  24. 2018Multi-Reference Training with Pseudo-References for Neural Translation and Text Generationcited by 15

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 15; 2026: 25.

40 papers.

Sort:

By area:

  1. 2025Memory in the Age of AI Agentscited by 303
  2. 2025DeepAgent: A General Reasoning Agent with Scalable Toolsetscited by 71
  3. 2025Tongyi DeepResearch Technical Reportcited by 66
  4. 2025Agentic Entropy-Balanced Policy Optimizationcited by 40
  5. 2026OpenSeeker: Democratizing Frontier Search Agents by Fully Open-Sourcing Training Datacited by 39
  6. 2026MiroThinker-1.7 & H1: Towards Heavy-Duty Research Agents via Verificationcited by 32
  7. 2026Qwen-AgentWorld: Language World Models for General Agentscited by 28
  8. 2025A Survey on Agentic Multimodal Large Language Modelscited by 23
  9. 2025Enterprise Deep Research: Steerable Multi-Agent Deep Research for Enterprise Analyticscited by 20
  10. 2026Agentic Aggregation for Parallel Scaling of Long-Horizon Agentic Taskscited by 18
  11. 2025Toward Effective Tool-Integrated Reasoning via Self-Evolved Preference Learningcited by 17
  12. 2025AutoForge: Automated Environment Synthesis for Agentic Reinforcement Learningcited by 17
28 more
  1. 2026Position: Agentic Evolution is the Path to Evolving LLMscited by 16
  2. 2025ParallelMuse: Agentic Parallel Thinking for Deep Information Seekingcited by 14
  3. 2025APTBench: Benchmarking Agentic Potential of Base LLMs During Pre-Trainingcited by 11
  4. 2025Klear-AgentForge: Forging Agentic Intelligence through Posttraining Scalingcited by 9
  5. 2026AgentSwing: Adaptive Parallel Context Management Routing for Long-Horizon Web Agentscited by 9
  6. 2025RhinoInsight: Improving Deep Research through Control Mechanisms for Model Behavior and Contextcited by 8
  7. 2026TRACE: Trajectory-Aware Comprehensive Evaluation for Deep Research Agentscited by 8
  8. 2026Agentic Environment Engineering for Large Language Models: A Survey of Environment Modeling, Synthesis, Evaluation, and Applicationcited by 8
  9. 2025MindWatcher: Toward Smarter Multimodal Tool-Integrated Reasoningcited by 6
  10. 2026On Training Large Language Models for Long-Horizon Tasks: An Empirical Study of Horizon Lengthcited by 6
  11. 2026Agent-Omit: Adaptive Context Omission for Efficient LLM Agentscited by 5
  12. 2026Self-Improvement of Large Language Models: A Technical Overview and Future Outlookcited by 5
  13. 2026ET-Agent: Incentivizing Effective Tool-Integrated Reasoning Agent via Behavior Calibrationcited by 4
  14. 2025Learning to Reason as Action Abstractions with Scalable Mid-Training RLcited by 3
  15. 2026From P(y|x) to P(y): Investigating Reinforcement Learning in Pre-train Spacecited by 3
  16. 2026Internalizing the Future: A Unified Agentic Training Paradigm for World Model Planningcited by 3
  17. 2025Thinking by Doing: Building Efficient World Model Reasoning in LLMs via Multi-turn Interactioncited by 2
  18. 2026DuMate-DeepResearch: An Auditable Multi-Agent System with Recursive Search and Rubric-Grounded Reasoningcited by 2
  19. 2026State2State: Environment-Derived Mid-Training for LLM Agentscited by 2
  20. 2026Apodex 1.1: Scaling Agentic Intelligence for Complex Workcited by 2
  21. 2026MemAdapter: Fast Alignment across Agent Memory Paradigms via Generative Subgraph Retrievalcited by 1
  22. 2026Safactory: A Scalable Agentic Infrastructure for Training Trustworthy Autonomous Intelligencecited by 1
  23. 2026ToolAnchor: Anchoring Counterfactual Context to Boost Agentic Tool-use Capabilitycited by 1
  24. 2026MidTool: Mid-training Data Synthesis for Agentic Tool Usecited by 1
  25. 2026FrontierChallenge: Evaluating Scientific Workflow Completioncited by 1
  26. 2026ExpRL: Exploratory RL for LLM Mid-Trainingcited by 0
  27. 2026ZGCM-1: A Fully Open and Extremely Efficient Foundation Model for Math and Agentic Searchcited by 0
  28. 2026TrajLong: Co-Designing Agentic and Long-Context Supervision for Mid-Trainingcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024Offline Training of Language Model Agents with Functions as Learnable Weights
  2. 2025Adaptation of Agentic AI: A Survey of Post-Training, Memory, and Skills
  3. 2026SPT: Skills as Pre-Training Data for Agentic Language Models
  4. 2025AgentDistill: Training-Free Agent Distillation with Generalizable MCP Boxes
  5. 2026Mock Worlds, Real Skills: Building Small Agentic Language Models with Synthetic Tasks, Simulated Environments, and Rubric-Based Rewards
  6. 2025SFR-DeepResearch: Towards Effective Reinforcement Learning for Autonomously Reasoning Single Agents
  7. 2026AgentCPM-Explore: Realizing Long-Horizon Deep Exploration for Edge-Scale Agents
  8. 2025AgentRefine: Enhancing Agent Generalization through Refinement Tuning