mlsanityData to 7 Oct 2026

On Data Engineering for Scaling LLM Terminal Capabilities

arXiv 2602.21193 ↗First posted 24 Feb 2026cs.CLcited by 34

Not on the map. Why?

Abstract (arXiv)

Despite rapid recent progress in the terminal capabilities of large language models, the training data strategies behind state-of-the-art terminal agents remain largely undisclosed. We address this gap through a systematic study of data engineering practices for terminal agents, making two key contributions: (1) Terminal-Task-Gen, a lightweight synthetic task generation pipeline that supports seed-based and skill-based task construction, and (2) a comprehensive analysis of data and training strategies, including filtering, curriculum learning, long context training, and scaling behavior. Our pipeline yields Terminal-Corpus, a large-scale open-source dataset for terminal tasks. Using this dataset, we train Nemotron-Terminal, a family of models initialized from Qwen3(8B, 14B, 32B) that achieve substantial gains on Terminal-Bench 2.0: Nemotron-Terminal-8B improves from 2.5% to 13.0% Nemotron-Terminal-14B improves from 4.0% to 20.2%, and Nemotron-Terminal-32B improves from 3.4% to 27.4%, matching the performance of significantly larger models. To accelerate research in this domain, we open-source our model checkpoints and most of our synthetic datasets at https://huggingface.co/collections/nvidia/nemotron-terminal.

Key facts. Cited by 34 papers in the map (data through 7 Oct 2026). Builds on 17 of its 41 references that are in the map. First posted on arXiv 24 Feb 2026.

Builds on

17 of 41 references.

References · 17

  1. 2025Qwen3 Technical Reportcited by 8,195
  2. 2023SWE-bench: Can Language Models Resolve Real-World GitHub Issues?cited by 3,193
  3. 2024HybridFlow: A Flexible and Efficient RLHF Frameworkcited by 1,980
  4. 2024LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Codecited by 1,906
  5. 2025DeepSeek-V3.2: Pushing the Frontier of Open Large Language Modelscited by 755
  6. 2023WizardCoder: Empowering Code Large Language Models with Evol-Instructcited by 728
  7. 2023YaRN: Efficient Context Window Extension of Large Language Modelscited by 584
  8. 2026Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfacescited by 515
  9. 2024Magpie: Alignment Data Synthesis from Scratch by Prompting Aligned LLMs with Nothingcited by 341
  10. 2025SWE-smith: Scaling Data for Software Engineering Agentscited by 255
  11. 2025AIMO-2 Winning Solution: Building State-of-the-Art Mathematical Reasoning Models with OpenMathReasoning datasetcited by 165
  12. 2025SWE-rebench: An Automated Pipeline for Task Collection and Decontaminated Evaluation of Software Engineering Agentscited by 127
5 more
  1. 2025OpenCodeReasoning: Advancing Data Distillation for Competitive Codingcited by 106
  2. 2024AgentInstruct: Toward Generative Teaching with Agentic Flowscited by 88
  3. 2025SWE-Fixer: Training Open-Source LLMs for Effective and Efficient GitHub Issue Resolutioncited by 87
  4. 2025Nemotron-Cascade: Scaling Cascaded Reinforcement Learning for General-Purpose Reasoning Modelscited by 45
  5. 2024LAB: Large-Scale Alignment for ChatBotscited by 39

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 34.

34 papers.

Sort:

By area:

  1. 2026Nemotron 3 Super: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoningcited by 33
  2. 2026Tmax: A simple recipe for terminal agentscited by 24
  3. 2026CUA-Gym: Scaling Verifiable Training Environments and Tasks for Computer-Use Agentscited by 21
  4. 2026Toward Scalable Terminal Task Synthesis via Skill Graphscited by 16
  5. 2026Terminal-World: Scaling Terminal-Agent Environments via Agent Skillscited by 13
  6. 2026A Self-Evolving Framework for Efficient Terminal Agents via Observational Context Compressioncited by 12
  7. 2026Failure as a Process: An Anatomy of CLI Coding Agent Trajectoriescited by 11
  8. 2026TerminalWorld: Benchmarking Agents on Real-World Terminal Taskscited by 10
  9. 2026Agentic Environment Engineering for Large Language Models: A Survey of Environment Modeling, Synthesis, Evaluation, and Applicationcited by 8
  10. 2026Recursive Synthesis for Long-Horizon Terminal Taskscited by 8
  11. 2026LiteCoder-Terminal: Scaling Long-Horizon Terminal Environments for Learning Language Agentscited by 7
  12. 2026CLI-Universe: Towards Verifiable Task Synthesis Engine for Terminal Agentscited by 7
22 more
  1. 2026What Makes Interaction Trajectories Effective for Training Terminal Agents?cited by 6
  2. 2026SPADE: Self-Play in Adaptive Synthetic Executable Environmentscited by 6
  3. 2026SETA: Scaling Environments for Terminal Agentscited by 5
  4. 2026CalibForge: Adversarial Solver Calibration for Scaling Learnable Terminal Taskscited by 5
  5. 2026NexForge: Scaling Agent Capabilities through Requirement-Driven Task Synthesis for LLMscited by 4
  6. 2026Intern-S2-Preview: Scientific Agentic Foundation Modelcited by 3
  7. 2026FACET: Preserving Source Intent and Executable State in Terminal Task Synthesiscited by 3
  8. 2026Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environmentscited by 3
  9. 2026Environment Evolution for Terminal Agentscited by 2
  10. 2026ModularRSI: Modular and Generalizable Recursive Harness Self-Improvementcited by 2
  11. 2026OPD-Evolver: Cultivating Holistic Agent Evolver via On-Policy Distillationcited by 1
  12. 2026Meta-Task: Turning Terminal Task Synthesis into a Terminal Task for Scalable Agent Trainingcited by 1
  13. 2026OctoLong: Mid-Training On Cross-Repository Code Contexts Enhances Long-Context Modelingcited by 1
  14. 2026Learning Generalizable Behaviors for Terminal Agentscited by 1
  15. 2026Jupiter-N Technical Reportcited by 0
  16. 2026unix-ctf: Procedural Environments for Unix-Competence Reinforcement Learningcited by 0
  17. 2026Post-training is (Massive) Supervised Learningcited by 0
  18. 2026Terminal Agents: A Survey of AI Agents in Command-Line Environmentscited by 0
  19. 2026Don't Mask the Environment: Observation Supervision Changes How Agents Explore Under RLcited by 0
  20. 2026SkillGym: Internalizing Human Skills into LLMs for Real-World Problem Solvingcited by 0
  21. 2026Rufus-Air: An Open LLM Post-Training Recipecited by 0
  22. 2026Mid-Harness: Scaling Actions Between Model and Harness for Terminal Agentscited by 0