mlsanityData to 14 Oct 2026

Synthesizing Post-Training Data for LLMs through Multi-Agent Simulation

arXiv 2410.14251 ↗First posted 18 Oct 2024cs.AI cs.CLcited by 24

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

Abstract (arXiv)

Post-training is essential for enabling large language models (LLMs) to follow human instructions. However, its effectiveness depends on high-quality instruction data, which is challenging to obtain in the real world due to privacy concerns, data scarcity, and high annotation costs. To fill this gap, inspired by the recent success of using LLMs to simulate human society, we propose MATRIX, a multi-agent simulator that automatically generates diverse text-based scenarios, capturing a wide range of real-world human needs in a realistic and scalable manner. Leveraging these outputs, we introduce a novel scenario-driven instruction generator MATRIX-Gen for controllable and highly realistic data synthesis. Extensive experiments demonstrate that our framework effectively generates both general and domain-specific data. On AlpacaEval 2 and Arena-Hard benchmarks, Llama-3-8B-Base, post-trained on datasets synthesized by MATRIX-Gen with just 20K instruction-response pairs, outperforms Meta's Llama-3-8B-Instruct model, which was trained on over 10M pairs.

Key facts. Cited by 24 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 48 of its 69 references that are in the map. First posted on arXiv 18 Oct 2024.

Builds on

48 of 69 references.

Closest ancestors · 2

Same direction, newest first

  1. 2024Scaling Large Language Model-based Multi-Agent Collaboration11 Jun
  2. 2023CAMEL: Communicative Agents for "Mind" Exploration of Large Language Model Society31 Mar

Other references · 46

  1. 2023GPT-4 Technical Reportcited by 18,396
  2. 2024The Llama 3 Herd of Modelscited by 14,715
  3. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  4. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,626
  5. 2021Evaluating Large Language Models Trained on Codecited by 8,320
  6. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 8,065
  7. 2020Measuring Massive Multitask Language Understandingcited by 7,963
  8. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,371
  9. 2023Let's Verify Step by Stepcited by 3,789
  10. 2022Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedbackcited by 3,720
  11. 2021Program Synthesis with Large Language Modelscited by 3,637
  12. 2023Generative Agents: Interactive Simulacra of Human Behaviorcited by 3,568
34 more
  1. 2023Universal and Transferable Adversarial Attacks on Aligned Language Modelscited by 3,104
  2. 2023GPQA: A Graduate-Level Google-Proof Q&A Benchmarkcited by 2,879
  3. 2022Self-Instruct: Aligning Language Models with Self-Generated Instructionscited by 2,724
  4. 2023Voyager: An Open-Ended Embodied Agent with Large Language Modelscited by 2,191
  5. 2024Qwen2 Technical Reportcited by 2,120
  6. 2023WizardLM: Empowering large pre-trained language models to follow complex instructionscited by 1,125
  7. 2023LIMA: Less Is More for Alignmentcited by 1,100
  8. 2024Length-Controlled AlpacaEval: A Simple Way to Debias Automatic Evaluatorscited by 897
  9. 2022Super-NaturalInstructions: Generalization via Declarative Instructions on 1600+ NLP Taskscited by 868
  10. 2023BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Datasetcited by 860
  11. 2023Enhancing Chat Language Models by Scaling High-quality Instructional Conversationscited by 839
  12. 2023OpenAssistant Conversations -- Democratizing Large Language Model Alignmentcited by 721
  13. 2021Cross-Task Generalization via Natural Language Crowdsourcing Instructionscited by 699
  14. 2024WildChat: 1M ChatGPT Interaction Logs in the Wildcited by 629
  15. 2022Using Large Language Models to Simulate Multiple Humans and Replicate Human Subject Studiescited by 515
  16. 2024From Crowdsourced Data to High-Quality Benchmarks: Arena-Hard and BenchBuilder Pipelinecited by 500
  17. 2024Scaling Synthetic Data Creation with 1,000,000,000 Personascited by 417
  18. 2023UltraFeedback: Boosting Language Models with Scaled AI Feedbackcited by 406
  19. 2023Orca: Progressive Learning from Complex Explanation Traces of GPT-4cited by 379
  20. 2022Text and Code Embeddings by Contrastive Pre-Trainingcited by 361
  21. 2024Magpie: Alignment Data Synthesis from Scratch by Prompting Aligned LLMs with Nothingcited by 346
  22. 2022One Embedder, Any Task: Instruction-Finetuned Text Embeddingscited by 333
  23. 2023Magicoder: Empowering Code Generation with OSS-Instructcited by 256
  24. 2024MT-Bench-101: A Fine-Grained Benchmark for Evaluating Large Language Models in Multi-Turn Dialoguescited by 250
  25. 2024OpenCodeInterpreter: Integrating Code Generation with Execution and Refinementcited by 247
  26. 2023Camels in a Changing Climate: Enhancing LM Adaptation with Tulu 2cited by 229
  27. 2023Self-Alignment with Instruction Backtranslationcited by 151
  28. 2024Best Practices and Lessons Learned on Synthetic Datacited by 106
  29. 2023AgentSims: An Open-Source Sandbox for Large Language Model Evaluationcited by 88
  30. 2024Unveiling the Truth and Facilitating Change: Towards Agent-based Large-scale Social Movement Simulationcited by 71
  31. 2024Can Large Language Model Agents Simulate Human Trust Behavior?cited by 64
  32. 2023Estimating Contamination via Perplexity: Quantifying Memorisation in Language Model Evaluationcited by 48
  33. 2024Self-Alignment of Large Language Models via Monopolylogue-based Social Scene Simulationcited by 46
  34. 2024Privacy-Preserving Instructions for Aligning Large Language Modelscited by 35

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2024: 1; 2025: 14; 2026: 9.

24 papers.

Sort:

By area:

  1. 2025APIGen-MT: Agentic Pipeline for Multi-Turn Data Generation via Simulated Agent-Human Interplaycited by 167
  2. 2024From Individual to Society: A Survey on Social Simulation Driven by Large Language Model-based Agentscited by 82
  3. 2025LoopTool: Closing the Data-Training Loop for Robust LLM Tool Callscited by 17
  4. 2025MASLab: A Unified and Comprehensive Codebase for LLM-based Multi-Agent Systemscited by 14
  5. 2025MetaSynth: Meta-Prompting-Driven Agentic Scaffolds for Diverse Synthetic Data Generationcited by 10
  6. 2025Behavior Injection: Preparing Language Models for Reinforcement Learningcited by 8
  7. 2026Towards On-Policy Data Evolution for Visual-Native Multimodal Deep Search Agentscited by 8
  8. 2025XtraGPT: Context-Aware and Controllable Academic Paper Revision via Human-AI Collaborationcited by 7
  9. 2025On the Importance of Task Complexity in Evaluating LLM-Based Multi-Agent Systemscited by 6
  10. 2025Combatting Dimensional Collapse in LLM Pre-Training Data via Diversified File Selectioncited by 5
  11. 2026Beyond Quantity: Trajectory Diversity Scaling for Code Agentscited by 5
  12. 2026TopoCurate:Modeling Interaction Topology for Tool-Use Agent Trainingcited by 4
12 more
  1. 2025Bohdi: Heterogeneous LLM Fusion with Automatic Data Explorationcited by 3
  2. 2025Beyond Syntax: Action Semantics Learning for App Agentscited by 3
  3. 2025FURINA: A Fully Customizable Role-Playing Benchmark via Scalable Multi-Agent Collaboration Pipelinecited by 3
  4. 2025EmboMatrix: A Scalable Training-Ground for Embodied Decision-Makingcited by 3
  5. 2025Refine-n-Judge: Curating High-Quality Preference Chains for LLM-Fine-Tuningcited by 2
  6. 2025Beyond Query-Level Comparison: Fine-Grained Reinforcement Learning for Text-to-SQL with Automated Interpretable Critiquescited by 2
  7. 2026MathAgent: Adversarial Evolution of Constraint Graphs for Mathematical Reasoning Data Synthesiscited by 2
  8. 2026Declarative by Design, Assistable Only by Convention: Benchmarking Multi-Agent Frameworks for AI-Assistabilitycited by 0
  9. 2026Know When To Fold 'Em: Token-Efficient LLM Synthetic Data Generation via Multi-Stage In-Flight Rejectioncited by 0
  10. 2026Better Literary Translation: A Multi-Aspect Data Generation and LLM Training Approachcited by 0
  11. 2026Playing social deduction games with reinforcement fine-tuned large language modelscited by 0
  12. 2026Know What You Don't Know: Agentic Data Synthesis by Evolving Generation Space for Closed-Loop Learningcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024Star-Agents: Automatic Data Optimization with LLM Agents for Instruction Tuning
  2. 2024TeaMs-RL: Teaching LLMs to Generate Better Instruction Datasets via Reinforcement Learning
  3. 2025LaMDAgent: An Autonomous Framework for Post-Training Pipeline Optimization via LLM Agents
  4. 2023AgentTuning: Enabling Generalized Agent Abilities for LLMs
  5. 2024AgentInstruct: Toward Generative Teaching with Agentic Flows
  6. 2026LLM Agents for Automated Post-Training via Structured Recipe Optimization
  7. 2025Language Modeling by Language Models
  8. 2026Enhancing LLM Instruction Following: An Evaluation-Driven Multi-Agentic Workflow for Prompt Instructions Optimization