mlsanityData to 14 Oct 2026

Evaluating Language Models as Synthetic Data Generators

arXiv 2412.03679 ↗First posted 4 Dec 2024cs.CLcited by 33

On the map: Synthetic data and self-improvement for LLMs → LLM synthetic data generation

Abstract (arXiv)

Given the increasing use of synthetic data in language model (LM) post-training, an LM's ability to generate high-quality data has become nearly as crucial as its ability to solve problems directly. While prior works have focused on developing effective data generation methods, they lack systematic comparison of different LMs as data generators in a unified setting. To address this gap, we propose AgoraBench, a benchmark that provides standardized settings and metrics to evaluate LMs' data generation abilities. Through synthesizing 1.26 million training instances using 6 LMs and training 99 student models, we uncover key insights about LMs' data generation capabilities. First, we observe that LMs exhibit distinct strengths. For instance, GPT-4o excels at generating new problems, while Claude-3.5-Sonnet performs better at enhancing existing ones. Furthermore, our analysis reveals that an LM's data generation ability doesn't necessarily correlate with its problem-solving ability. Instead, multiple intrinsic features of data quality-including response quality, perplexity, and instruction difficulty-collectively serve as better indicators. Finally, we demonstrate that strategic choices in output format and cost-conscious model selection significantly impact data generation effectiveness.

Key facts. Cited by 33 papers in the map (data through 14 Oct 2026). Placed in LLM synthetic data generation, one of 13 directions in Synthetic data and self-improvement for LLMs. Builds on 39 of its 49 references that are in the map. First posted on arXiv 4 Dec 2024.

Builds on

39 of 49 references.

References · 39

  1. 2020Language Models are Few-Shot Learnerscited by 32,419
  2. 2024The Llama 3 Herd of Modelscited by 14,715
  3. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  4. 2021Evaluating Large Language Models Trained on Codecited by 8,320
  5. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,371
  6. 2024GPT-4o System Cardcited by 4,588
  7. 2021Finetuned Language Models Are Zero-Shot Learnerscited by 3,780
  8. 2021Program Synthesis with Large Language Modelscited by 3,637
  9. 2022Self-Instruct: Aligning Language Models with Self-Generated Instructionscited by 2,724
  10. 2023Is Your Code Generated by ChatGPT Really Correct? Rigorous Evaluation of Large Language Models for Code Generationcited by 1,391
  11. 2023A Survey on Evaluation of Large Language Modelscited by 1,242
  12. 2023WizardLM: Empowering large pre-trained language models to follow complex instructionscited by 1,125
27 more
  1. 2022Super-NaturalInstructions: Generalization via Declarative Instructions on 1600+ NLP Taskscited by 868
  2. 2023The Flan Collection: Designing Data and Methods for Effective Instruction Tuningcited by 739
  3. 2021Cross-Task Generalization via Natural Language Crowdsourcing Instructionscited by 699
  4. 2022MTEB: Massive Text Embedding Benchmarkcited by 640
  5. 2023Textbooks Are All You Needcited by 569
  6. 2024From Crowdsourced Data to High-Quality Benchmarks: Arena-Hard and BenchBuilder Pipelinecited by 500
  7. 2024RewardBench: Evaluating Reward Models for Language Modelingcited by 499
  8. 2023MAmmoTH: Building Math Generalist Models through Hybrid Instruction Tuningcited by 487
  9. 2024Prometheus 2: An Open Source Language Model Specialized in Evaluating Other Language Modelscited by 446
  10. 2022Unnatural Instructions: Tuning Language Models with (Almost) No Human Laborcited by 397
  11. 2023Orca: Progressive Learning from Complex Explanation Traces of GPT-4cited by 379
  12. 2023What Makes Good Data for Alignment? A Comprehensive Study of Automatic Data Selection in Instruction Tuningcited by 364
  13. 2024Magpie: Alignment Data Synthesis from Scratch by Prompting Aligned LLMs with Nothingcited by 346
  14. 2024Skywork-Reward: Bag of Tricks for Reward Modeling in LLMscited by 327
  15. 2023How Abilities in Large Language Models are Affected by Supervised Fine-tuning Data Compositioncited by 223
  16. 2023Evaluating Large Language Models: A Comprehensive Surveycited by 202
  17. 2018Learning to Mine Aligned Code and Natural Language Pairs from Stack Overflowcited by 165
  18. 2022Instruction Induction: From Few Examples to Natural Language Task Descriptionscited by 152
  19. 2024MAmmoTH2: Scaling Instructions from the Webcited by 145
  20. 2024Observational Scaling Laws and the Predictability of Language Model Performancecited by 130
  21. 2023The CoT Collection: Improving Zero-shot and Few-shot Learning of Language Models via Chain-of-Thought Fine-Tuningcited by 108
  22. 2024MixEval: Deriving Wisdom of the Crowd from LLM Benchmark Mixturescited by 74
  23. 2023Prompt2Model: Generating Deployable Models from Natural Language Instructionscited by 41
  24. 2024Unveiling the Impact of Coding Data Instruction Fine-Tuning on Large Language Models Reasoningcited by 31
  25. 2024Stronger Models are NOT Stronger Teachers for Instruction Tuningcited by 16
  26. 2023CoTEVer: Chain of Thought Prompting Annotation Toolkit for Explanation Verificationcited by 11
  27. 2024Universal and Context-Independent Triggers for Precise Control of LLM Outputscited by 4

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 21; 2026: 12.

33 papers.

Sort:

By area:

  1. 2025Small Models Struggle to Learn from Strong Reasonerscited by 113
  2. 2025EvalTree: Profiling Language Model Weaknesses via Hierarchical Capability Treescited by 27
  3. 2025ALFA: Aligning LLMs to Ask Good Questions A Case Study in Clinical Reasoningcited by 25
  4. 2025Find Your Optimal Teacher: Personalized Data Synthesis via Router-Guided Multi-Teacher Distillationcited by 15
  5. 2025Optimizing Diversity and Quality through Base-Aligned Model Collaborationcited by 8
  6. 2025MT-RAIG: Novel Benchmark and Evaluation Framework for Retrieval-Augmented Insight Generation over Multiple Tablescited by 7
  7. 2025RECAP: REwriting Conversations for Intent Understanding in Agentic Planningcited by 7
  8. 2025Attributes as Textual Genes: Leveraging LLMs as Genetic Algorithm Simulators for Conditional Synthetic Data Generationcited by 5
  9. 2026Pythagoras-Prover: Advancing Efficient Formal Proving via Augmented Lean Formalisationcited by 4
  10. 2025Data Swarms: Optimizable Generation of Synthetic Evaluation Datacited by 2
  11. 2025FLAMES: Improving LLM Math Reasoning via a Fine-Grained Analysis of the Data Synthesis Pipelinecited by 2
  12. 2025Generative Models for Synthetic Data: Transforming Data Mining in the GenAI Eracited by 2
21 more
  1. 2025TiTok: Transfer Token-level Knowledge via Contrastive Excess to Transplant LoRAcited by 2
  2. 2026Exploring Autonomous Agentic Data Engineering for Model Specializationcited by 2
  3. 2025Efficient Response Generation Strategy Selection for Fine-Tuning Large Language Models Through Self-Aligned Perplexitycited by 1
  4. 2025Modular Techniques for Synthetic Long-Context Data Generation in Language Model Training and Evaluationcited by 1
  5. 2025RoSE: Round-robin Synthetic Data Evaluation for Selecting LLM Generators without Human Test Setscited by 1
  6. 2025Performance Trade-offs of Optimizing Small Language Models for E-Commercecited by 1
  7. 2025Classifying German Language Proficiency Levels Using Large Language Modelscited by 1
  8. 2026ExStrucTiny: A Benchmark for Schema-Variable Structured Information Extraction from Document Imagescited by 1
  9. 2026Polyglot Teachers: Evaluating Language Models for Multilingual Synthetic Data Generationcited by 1
  10. 2026AcquisitionSynthesis: Targeted Data Generation using Acquisition Functionscited by 1
  11. 2025Enhancing LLMs for Identifying and Prioritizing Important Medical Jargons from Electronic Health Record Notes Utilizing Data Augmentation: A Comparative Studycited by 0
  12. 2025Self-Bootstrapping Automated Program Repair: Using LLMs to Generate and Evaluate Synthetic Training Data for Bug Repaircited by 0
  13. 2025Spatial Knowledge Graph-Guided Multimodal Synthesiscited by 0
  14. 2025TextMineX: Data, Evaluation Framework and Ontology-guided LLM Pipeline for Humanitarian Mine Actioncited by 0
  15. 2026SeaAlert: Robust Severity Classification and LLM-Based Information Extraction for Noisy Maritime Distress Communicationscited by 0
  16. 2026NodeSynth: Socially Aligned Synthetic Data for AI Evaluationcited by 0
  17. 2026ROBE: Reversed-Order-Biased-Experts for Extracting Extreme Long-tail Events from Historical Textscited by 0
  18. 2026ToolLoop: Closed-Loop Tool-Use Data Synthesis via Decomposed Generation and Dynamic Self-Feedbackcited by 0
  19. 2026AMU:Admission and Memory Update for Personalized Conversations---Structured Memory with SLM Guided Controlcited by 0
  20. 2026Synthetic Data Characterization via Training Dynamicscited by 0
  21. 2026Metric-Construction Coupling Inflates Measured Synthetic Dialect Recoverycited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Towards Active Synthetic Data Generation for Finetuning Language Models
  2. 2024A Survey on Data Synthesis and Augmentation for Large Language Models
  3. 2025Synthetic Data Generation Using Large Language Models: Advances in Text and Code
  4. 2026A Survey on Evaluating Quality and Trustworthiness in LLM-Generated Data
  5. 2024Towards a Theoretical Understanding of Synthetic Data in LLM Post-Training: A Reverse-Bottleneck Perspective
  6. 2024Balancing Cost and Effectiveness of Synthetic Data Generation Strategies for LLMs
  7. 2023Synthetic Data Generation with Large Language Models for Text Classification: Potential and Limitations
  8. 2025BARE: Leveraging Base Language Models for Few-Shot Synthetic Data Generation