Shadowgraph
TrendingAbout
Sign in

Egg-Hu/Awesome-Synthetic-Data-Generation

23

18 commits

updated Jan 7, 2026

See the code

README

Awesome Synthetic Data Generation

PRs Welcome Stars Forks

A comprehensive survey and curated collection of resources on synthetic data generation.

📚 Table of Contents (Click to Expand)

1. Methodologies

1.1 Generation-Based Synthesis

Synthesis from scratch

Paper TitleYearConference/Journal
Absolute zero: Reinforced self-play reasoning with zero data2025arXiv
Synthesizing post-training data for llms through multi-agent simulation2024arXiv
Learning vision from models rivals learning vision from data2024CVPR
Mathvista: Evaluating mathematical reasoning of foundation models in visual contexts2023arXiv
Dreamteacher: Pretraining image backbones with deep generative models2023ICCV
Self-instruct: Aligning language models with self-generated instructions2022arXiv
Training verifiers to solve math word problems2021arXiv

Synthesis from seeds

Paper TitleYearConference/Journal
Spread Preference Annotation: Direct Preference Judgment for Efficient LLM Alignment2025ICLR
Human-guided image generation for expanding small-scale training image datasets2025TVCG
Cream: Consistency regularized self-rewarding language models2024arXiv
Distribution-aware data expansion with diffusion models2024NeurIPS
Mitigating catastrophic forgetting in large language models with self-synthesized rehearsal2024arXiv
Towards automating text annotation: A case study on semantic proximity annotation using gpt-42024arXiv
Self-Rewarding Language Models2024arXiv
Real-fake: Effective training data synthesis through distribution matching2023arXiv
Expanding small-scale datasets with guided imagination2023NeurIPS
Training on thin air: Improve image classification with generated data2023arXiv

Synthesis from structure

Paper TitleYearConference/Journal
Synthesize-on-Graph: Knowledgeable Synthetic Data Generation for Continue Pre-training of Large Language Models2025arXiv
Logictree: Improving complex reasoning of LLMs via instantiated multi-step synthetic logical data2025NeurIPS
A Graph-Based Synthetic Data Pipeline for Scaling High-Quality Reasoning Instructions2024arXiv
MathScale: Scaling Instruction Tuning for Mathematical Reasoning2024ICML
ControlMath: Controllable Data Generation Promotes Math Generalist Models2024EMNLP
Enhancing logical reasoning in large language models through graph-based synthetic data2024arXiv
Infinitymath: A scalable instruction tuning dataset in programmatic mathematical reasoning2024CIKM
Synthetic data (almost) from scratch: Generalized instruction tuning for language models2024arXiv

Synthesis with evolution

Paper TitleYearConference/Journal
R-Zero: Self-Evolving Reasoning LLM from Zero Data2025arXiv
Prismatic Synthesis: Gradient-based Data Diversification Boosts Generalization in LLM Reasoning2025arXiv
HS-STAR: Hierarchical Sampling for Self-Taught Reasoners via Difficulty Estimation and Budget Reallocation2025arXiv
Enhancing LLM Reasoning with Iterative DPO: A Comprehensive Empirical Investigation2025arXiv
Spend wisely: Maximizing post-training gains in iterative synthetic data boostrapping2025arXiv
Scaling laws of synthetic images for model training... for now2024CVPR
Cream: Consistency regularized self-rewarding language models2024arXiv
Dualformer: Controllable fast and slow thinking by learning with randomized reasoning traces2024arXiv
Improving clip training with language rewrites2023NeurIPS
Feedback-guided data synthesis for imbalanced classification2023arXiv
Raft: Reward ranked finetuning for generative foundation model alignment2023arXiv
Synthetic data from diffusion models improves imagenet classification2023arXiv
Self-Rewarding Language Models2024arXiv
Spread Preference Annotation: Direct Preference Judgment for Efficient LLM Alignment2025ICLR
Star: Bootstrapping reasoning with reasoning2022NeurIPS
Unleashing reasoning capability of llms via scalable question synthesis from scratch2024arXiv
A Survey on LLM-as-a-Judge2024arXiv

1.2 Inversion-Based Synthesis

Data-space inversion

Paper TitleYearConference/Journal
Reverse-Engineered Reasoning for Open-Ended Generation2025arXiv
Dreaming to distill: Data-free knowledge transfer via deepinversion2020CVPR

Latent-space inversion

Paper TitleYearConference/Journal
Generative Model Inversion Through the Lens of the Manifold Hypothesis2025arXiv
Pseudo-private data guided model inversion attacks2024NeurIPS
Open-vocabulary customization from clip via data-free knowledge distillation2025ICLR
Reinforcement learning-based black-box model inversion attacks2023CVPR
Re-thinking model inversion attacks against deep neural networks2023CVPR
Learning to Learn from APIs: Black-Box Data-Free Meta-Learning2023arXiv
Label-only model inversion attacks via knowledge transfer2023NeurIPS
Label-only model inversion attacks via boundary repulsion2022CVPR
Mirror: Model inversion for deep learning network with high fidelity2022NDSS
Variational model inversion attacks2021NeurIPS
Knowledge-enriched distributional model inversion attacks2021ICCV
Contrastive model inversion for data-free knowledge distillation2021arXiv
Deep image prior2018CVPR
The secret revealer: Generative model-inversion attacks against deep neural networks2020CVPR
Plug & play attacks: Towards robust and flexible model inversion attacks2022arXiv

1.3 Simulation-Based Synthesis

Agent-based simulation

Paper TitleYearConference/Journal
Synthesizing post-training data for llms through multi-agent simulation2024arXiv
Autogen: Enabling next-gen LLM applications via multi-agent conversations2024COLM
Camel: Communicative agents for" mind" exploration of large language model society2023NeurIPS
MetaGPT: Meta programming for a multi-agent collaborative framework2023ICLR

Platform-based simulation

Paper TitleYearConference/Journal
Synthesizing post-training data for llms through multi-agent simulation2024arXiv
Autogen: Enabling next-gen LLM applications via multi-agent conversations2024COLM
Camel: Communicative agents for" mind" exploration of large language model society2023NeurIPS
MetaGPT: Meta programming for a multi-agent collaborative framework2023ICLR
Unity Perception: Generate Synthetic Data for Computer Vision2022arXiv
Omniverse Replicator: Synthetic Data Generation for AI2021arXiv

1.4 Augmentation-Based Synthesis

Rule-based augmentation

Paper TitleYearConference/Journal
Cutmix: Regularization strategy to train strong classifiers with localizable features2019ICCV
EDA: Easy data augmentation techniques for boosting performance on text classification tasks2019arXiv
mixup: Beyond empirical risk minimization2017arXiv

Generative augmentation

Paper TitleYearConference/Journal
Data augmentation for image classification using generative ai2025WACV
Self-improving diffusion models with synthetic data2024arXiv
On the diversity of synthetic data and its impact on training large language models2024arXiv
Scaling synthetic data creation with 1,000,000,000 personas2024arXiv
Synthetic continued pretraining2024arXiv
Dart-math: Difficulty-aware rejection tuning for mathematical problem-solving2024NeurIPS
Detdiffusion: Synergizing generative and perceptive models for enhanced data generation and perception2024CVPR
Diffusemix: Label-preserving data augmentation with diffusion models2024CVPR
Dreamda: Generative data augmentation with diffusion models2024arXiv
Learning to augment synthetic images for sim2real policy transfer2019IROS

2. Applications

2.1 Data-centric AI

Data Accessibility

Zero/Few-shot learning

Paper TitleYearConference/Journal
Absolute zero: Reinforced self-play reasoning with zero data2025arXiv
Generating synthetic datasets for few-shot prompt tuning2024arXiv
View-invariant policy learning via zero-shot novel view synthesis2024arXiv
Prompting-based synthetic data generation for few-shot question answering2024arXiv
Datadream: Few-shot guided dataset generation2024ECCV
Learning vision from models rivals learning vision from data2024CVPR
Tuning language models as training data generators for augmentation-enhanced few-shot learning2023PMLR

Federated learning

Paper TitleYearConference/Journal
Federated Knowledge Recycling: Privacy-preserving synthetic data sharing2025PATTERN RECOGN LETT
Synthetic data aided federated learning using foundation models2024arXiv
Understanding Data Reconstruction Leakage in Federated Learning from a Theoretical Perspective2024arXiv
Gradient inversion of federated diffusion models2024arXiv
VertiMRF: Differentially Private Vertical Federated Data Synthesis2024KDD
Data-Free Federated Class Incremental Learning with Diffusion-Based Generative Memory2024arXiv
Fine-tuning global model via data-free knowledge distillation for non-iid federated learning2022CVPR
DENSE: Data-Free One-Shot Federated Learning2022NeurIPS
Parameterized Data-free knowledge distillation for heterogeneous federated learning2021ICML

Data-free knowledge distillation

Paper TitleYearConference/Journal
Open-vocabulary customization from clip via data-free knowledge distillation2025ICLR
Synthetic image learning: Preserving performance and preventing membership inference attacks2025PATTERN RECOGN LETT
Aligning teacher with student preferences for tailored training data generation2024arXiv
Model Conversion via Differentially Private Data-Free Distillation2023arXiv
Data-Free Knowledge Distillation via Feature Exchange and Activation Region Constraint2023CVPR
Re-thinking model inversion attacks against deep neural networks2023CVPR
Up to 100x faster data-free knowledge distillation2022AAAI
Momentum Adversarial Distillation: Handling Large Distribution Shifts in Data-Free Knowledge Distillation2022NeurIPS
Dreaming to distill: Data-free knowledge transfer via deepinversion2020CVPR
Data-free knowledge distillation with soft targeted transfer set synthesis2021AAAI
Contrastive model inversion for data-free knowledge distillation2021arXiv
Data-free knowledge distillation for object detection2021WACV
Data-free learning of student networks2019ICCV

Data-free pruning/quantization

Paper TitleYearConference/Journal
Sharpness-aware data generation for zero-shot quantization2024arXiv
Distilled Pruning: Using Synthetic Data to Win the Lottery2023arXiv

Data-free meta-learning

Paper TitleYearConference/Journal
Task-distributionally robust data-free meta-learning2025TPAMI
FREE: Faster and Better Data-Free Meta-Learning2024CVPR
Meta-learning without data via unconditional diffusion models2024TCSVT
Task Groupings Regularization: Data-Free Meta-Learning with Heterogeneous Pre-trained Models2024arXiv
Architecture, Dataset and Model-Scale Agnostic Data-free Meta-Learning2023CVPR
Learning to Learn from APIs: Black-Box Data-Free Meta-Learning2023arXiv

Data-free continual learning

Paper TitleYearConference/Journal
Data-Free Federated Class Incremental Learning with Diffusion-Based Generative Memory2024arXiv
Mitigating catastrophic forgetting in large language models with self-synthesized rehearsal2024arXiv
Self-distillation bridges distribution gap in language model fine-tuning2024arXiv
A data-free approach to mitigate catastrophic forgetting in federated class incremental learning for vision tasks2023NeurIPS
Ddgr: Continual learning with deep diffusion-based generative replay2023ICML

Data Refinement

Dataset distillation

Paper TitleYearConference/Journal
Unlocking Dataset Distillation with Diffusion Models2025NeurIPS
Generative dataset distillation based on diffusion model2024ECCV
Dim: Distilling dataset into generative model2024ECCV
Dataset-distillation generative model for speech emotion recognition2024arXiv
Generalizing dataset distillation via deep generative prior2023CVPR
Dataset condensation with distribution matching2023PMLR
Dataset distillation using neural feature regression2022NeurIPS
Dataset distillation by matching training trajectories2022CVPR
Cafe: Learning to condense dataset by aligning features2022CVPR
Dataset condensation with gradient matching2020arXiv
Dataset distillation2018arXiv

Dataset purification

Paper TitleYearConference/Journal
Adbm: Adversarial diffusion bridge model for reliable adversarial purification2024arXiv
Better synthetic data by retrieving and transforming existing datasets2024arXiv
Dataelixir: Purifying poisoned dataset to mitigate backdoor attacks via diffusion models2024AAAI
Diffusion models for adversarial purification2022arXiv

2.2 Model-centric AI

General Model Enhancement

General ability

Paper TitleYearConference/Journal
Synthesize-on-Graph: Knowledgeable Synthetic Data Generation for Continue Pre-training of Large Language Models2025arXiv
Large language models, physics-based modeling, experimental measurements: the trinity of data-scarce learning of polymer properties2024arXiv
Nemotron-CC: Transforming Common Crawl into a refined long-horizon pretraining dataset2024arXiv
Rephrasing the web: A recipe for compute and data-efficient language modeling2024arXiv
Scilitllm: How to adapt llms for scientific literature understanding2024arXiv
Synthetic continued pretraining2024arXiv
VILA $^ 2$: VILA Augmented VILA2024arXiv
Fine-tuning or retrieval? comparing knowledge injection in llms2023arXiv
Llemma: An open language model for mathematics2023arXiv
Textbooks are all you need2023arXiv
Textbooks are all you need ii: phi-1.5 technical report2023arXiv
Tinystories: How small can language models be and still speak coherent english?2023arXiv
Enhancing multilingual language model with massive multilingual knowledge triples2021arXiv

Domain Model Enhancement

Reasoning

Paper TitleYearConference/Journal
Absolute zero: Reinforced self-play reasoning with zero data2025arXiv
HS-STAR: Hierarchical Sampling for Self-Taught Reasoners via Difficulty Estimation and Budget Reallocation2025arXiv
Learning to Pose Problems: Reasoning-Driven and Solver-Adaptive Data Synthesis for Large Reasoning Models2025arXiv
Logictree: Improving complex reasoning of LLMs via instantiated multi-step synthetic logical data2025NeurIPS
Prismatic Synthesis: Gradient-based Data Diversification Boosts Generalization in LLM Reasoning2025arXiv
Seed-Coder: Let the Code Model Curate Data for Itself2025arXiv
Spread Preference Annotation: Direct Preference Judgment for Efficient LLM Alignment2025ICLR
Synthesize-on-Graph: Knowledgeable Synthetic Data Generation for Continue Pre-training of Large Language Models2025arXiv
Thinking LLMs: General Instruction Following with Thought Generation2025ICML
Unleashing Reasoning Capability of LLMs via Scalable Question Synthesis from Scratch2025arXiv
A Graph-Based Synthetic Data Pipeline for Scaling High-Quality Reasoning Instructions2024arXiv
Aligning teacher with student preferences for tailored training data generation2024arXiv
Autocoder: Enhancing code large language model with$\backslash$textsc $$AIEV-Instruct$$2024arXiv
Boosting reward model with preference-conditional multi-aspect synthetic data generation2024arXiv
ControlMath: Controllable Data Generation Promotes Math Generalist Models2024EMNLP
From the Least to the Most: Building a Plug-and-Play Visual Reasoner via Data Synthesis2024EMNLP
HexaCoder: Secure Code Generation via Oracle-Guided Synthetic Training Data2024arXiv
Infinitymath: A scalable instruction tuning dataset in programmatic mathematical reasoning2024CIKM
Jiuzhang3. 0: Efficiently improving mathematical reasoning by training small data synthesis models2024NeurIPS
MAmmoTH: Building Math Generalist Models through Hybrid Instruction Tuning2024ICLR
Marco-o1: Towards open reasoning models for open-ended solutions2024arXiv
MathScale: Scaling Instruction Tuning for Mathematical Reasoning2024ICML
MetaMath: Bootstrap Your Own Mathematical Questions for Large Language Models2024ICLR
Openmathinstruct-1: A 1.8 million math instruction tuning dataset2024NeurIPS
Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking2024COLM
Refined direct preference optimization with synthetic data for behavioral alignment of llms2024arXiv
Rl on incorrect synthetic data scales the efficiency of llm math reasoning by eight-fold2024NeurIPS
Self-Consistency Preference Optimization2024arXiv
Self-play with execution feedback: Improving instruction-following capabilities of large language models2024arXiv
Self-Rewarding Language Models2024arXiv
Small Language Models Need Strong Verifiers to Self-Correct Reasoning2024ACL
Strengthening multimodal large language model with bootstrapped preference optimization2024ECCV
ToRA: A Tool-Integrated Reasoning Agent for Mathematical Problem Solving2024ICLR
Tree-instruct: A preliminary study of the intrinsic relationship between complexity and alignment2024COLING
WizardLM: Empowering large pre-trained language models to follow complex instructions2024ICLR
Lamini-lm: A diverse herd of distilled models from large-scale instructions2023arXiv
Mathcoder: Seamless code integration in llms for enhanced mathematical reasoning2023arXiv
Reflection-tuning: Recycling data for better instruction-tuning2023NeurIPS
What makes good data for alignment? a comprehensive study of automatic data selection in instruction tuning2023arXiv
Wizardcoder: Empowering code large language models with evol-instruct2023arXiv
Wizardmath: Empowering mathematical reasoning for large language models via reinforced evol-instruct2023arXiv
Star: Bootstrapping reasoning with reasoning2022NeurIPS

Code

Paper TitleYearConference/Journal
Absolute zero: Reinforced self-play reasoning with zero data2025arXiv
Case2Code: Scalable Synthetic Data for Code Generation2025COLING
Increasing LLM Coding Capabilities through Diverse Synthetic Coding Tasks2025arXiv
Seed-Coder: Let the Code Model Curate Data for Itself2025arXiv
Autocoder: Enhancing code large language model with$\backslash$textsc $$AIEV-Instruct$$2024arXiv
Codeclm: Aligning language models with tailored synthetic data2024arXiv
HexaCoder: Secure Code Generation via Oracle-Guided Synthetic Training Data2024arXiv
WizardLM: Empowering large pre-trained language models to follow complex instructions2024ICLR
Wizardcoder: Empowering code large language models with evol-instruct2023arXiv

Instruction following

Paper TitleYearConference/Journal
Recursive introspection: Teaching language model agents how to self-improve2024NeurIPS
Selective reflection-tuning: Student-selected data recycling for llm instruction-tuning2024ACL
Self-play with execution feedback: Improving instruction-following capabilities of large language models2024arXiv
Self-refine instruction-tuning for aligning reasoning in language models2024arXiv
Spar: Self-play with tree-search refinement to improve instruction-following in large language models2024arXiv
Tree-instruct: A preliminary study of the intrinsic relationship between complexity and alignment2024LREC-COLING
WizardLM: Empowering large pre-trained language models to follow complex instructions2024ICLR
# instag: Instruction tagging for analyzing supervised fine-tuning of large language models2023arXiv
Alpagasus: Training a better alpaca with fewer data2023arXiv
Lamini-lm: A diverse herd of distilled models from large-scale instructions2023arXiv
Lima: Less is more for alignment2023NeurIPS
Reflection-tuning: Recycling data for better instruction-tuning2023NeurIPS
What makes good data for alignment? a comprehensive study of automatic data selection in instruction tuning2023arXiv
Self-instruct: Aligning language models with self-generated instructions2022arXiv

Preference

Paper TitleYearConference/Journal
Spread Preference Annotation: Direct Preference Judgment for Efficient LLM Alignment2025ICLR
Aligning teacher with student preferences for tailored training data generation2024arXiv
Boosting reward model with preference-conditional multi-aspect synthetic data generation2024arXiv
Course-correction: Safety alignment using synthetic preferences2024arXiv
Refined direct preference optimization with synthetic data for behavioral alignment of llms2024arXiv
Self-Consistency Preference Optimization2024arXiv
Self-directed synthetic dialogues and revisions technical report2024arXiv
Strengthening multimodal large language model with bootstrapped preference optimization2024ECCV

In-context learning

Paper TitleYearConference/Journal
Embracing Collaboration Over Competition: Condensing Multiple Prompts for Visual In-Context Learning2025CVPR
Self-Harmonized Chain of Thought2025arXiv
Auto-ICL: In-Context Learning without Human Supervision2024arXiv
Demonstration Augmentation for Zero-shot In-context Learning2024ACL
Self-Prompting Large Language Models for Zero-Shot Open-Domain QA2024arXiv
Automatic Chain of Thought Prompting in Large Language Models2023ICLR
Generate Rather than Retrieve: Large Language Models Are Strong Context Generators2023arXiv
Privacy-preserving in-context learning with differentially private few-shot generation2023arXiv
Recitation-Augmented Language Models2023arXiv
Self-ICL: Zero-Shot In-Context Learning with Self-Generated Demonstrations2023EMNLP
Self-Generated In-Context Learning: Leveraging Auto-regressive Language Models as a Demonstration Generator2022arXiv

Reinforcement Learning

Paper TitleYearConference/Journal
Kimi K2: Open Agentic Intelligence2025arXiv
s1: Simple test-time scaling2025arXiv
Synthetic Data Generation & Multi-Step RL for Reasoning & Tool Use2025arXiv
Synthetic Data RL: Task Definition Is All You Need2025arXiv
RL on Incorrect Synthetic Data Scales the Efficiency of LLM Math Reasoning by Eight-Fold2024arXiv

Model Evaluation

Synthetic benchmark

Paper TitleYearConference/Journal
Achilles' Heel of Mamba: Essential difficulties of the Mamba architecture demonstrated by synthetic data2025arXiv
Measuring General Intelligence with Generated Games2025arXiv
Efficacy of Synthetic Data as a Benchmark2024arXiv
Can you rely on your model evaluation? improving model evaluation with synthetic test data2023NeurIPS
SynBench: Task-Agnostic Benchmarking of Pretrained Representations using Synthetic Data2022NeurIPS

2.3 Trustworthy AI

Privacy

Privacy-preserving learning

Paper TitleYearConference/Journal
Does Training with Synthetic Data Truly Protect Privacy?2025arXiv
Federated Knowledge Recycling: Privacy-preserving synthetic data sharing2025PATTERN RECOGN LETT
Synthetic image learning: Preserving performance and preventing membership inference attacks2025PATTERN RECOGN LETT
Differentially Private Synthetic Data via Foundation Model APIs 1: Images)2024ICLR
Differentially private synthetic data via foundation model APIs 2: Text2024arXiv
Enhancing the utility of privacy-preserving cancer classification using synthetic data2024MICCAI
Generate synthetic text approximating the private distribution with differential privacy2024IJCAI
Privacy-preserving instructions for aligning large language models2024arXiv
SETSUBUN: Revisiting Membership Inference Game for Evaluating Synthetic Data Generation2024J INF PROCESS SYST
Synthetic data aided federated learning using foundation models2024arXiv
Towards Privacy-Preserving Relational Data Synthesis via Probabilistic Relational Models2024German Conference on Artificial Intellig
Using synthetic data to mitigate unfairness and preserve privacy through single-shot federated learning2024arXiv
VertiMRF: Differentially Private Vertical Federated Data Synthesis2024KDD

Model inversion attack

Paper TitleYearConference/Journal
Gradient inversion of federated diffusion models2024arXiv
Understanding Data Reconstruction Leakage in Federated Learning from a Theoretical Perspective2024arXiv
The secret revealer: Generative model-inversion attacks against deep neural networks2020CVPR
A methodology for formalizing model-inversion attacks2016CSF
Model inversion attacks that exploit confidence information and basic countermeasures2015SIGSAC

Safety & Security

Model stealing attack

Paper TitleYearConference/Journal
Unifying Multimodal Large Language Model Capabilities and Modalities via Model Merging2025arXiv
Data-free model extraction2021CVPR
Maze: Data-free model stealing attack using zeroth-order gradient estimation2021CVPR

Adversarial defense

Paper TitleYearConference/Journal
Deceptive diffusion: generating synthetic adversarial examples2025SSVM
Is synthetic data all we need? benchmarking the robustness of models trained with synthetic images2024CVPR
TSynD: Targeted Synthetic Data Generation for Enhanced Medical Image Classification: Leveraging Epistemic Uncertainty to Improve Model Performance2024arXiv
Leaving reality to imagination: Robust classification via generated datasets2023arXiv
Robust learning meets generative models: Can proxy distributions improve adversarial robustness?2021arXiv
Explaining and harnessing adversarial examples2014arXiv

Machine unlearning

Paper TitleYearConference/Journal
Mitigating catastrophic forgetting in large language models with self-synthesized rehearsal2024arXiv

Fairness

De-bias learning

Paper TitleYearConference/Journal
Generating Informative Samples for Risk-Averse Fine-Tuning of Downstream Tasks2025NeurIPS
Know" No''Better: A Data-Driven Approach for Enhancing Negation Awareness in CLIP2025arXiv
Modeling multi-task model merging as adaptive projective gradient descent2025ICML
Virus Infection Attack on LLMs: Your Poisoning Can Spread" VIA" Synthetic Data2025arXiv
FADE: Towards Fairness-aware Generation for Domain Generalization via Classifier-Guided Score-based Diffusion Models2024arXiv
Refined direct preference optimization with synthetic data for behavioral alignment of llms2024ICML
Strengthening multimodal large language model with bootstrapped preference optimization2024ECCV
SynthFair: Ensuring Subgroup Fairness in Classification via Synthetic Data Generation2024CSCE
Using synthetic data to mitigate unfairness and preserve privacy through single-shot federated learning2024arXiv
Discover and cure: Concept-aware mitigation of spurious correlation2023ICML

Long-tail learning

Paper TitleYearConference/Journal
DALK: Dynamic Co-Augmentation of LLMs and KG to answer Alzheimer's Disease Questions with Scientific Literature2024arXiv
Ltgc: Long-tail recognition via leveraging llms-driven generated content2024CVPR
Feedback-guided data synthesis for imbalanced classification2023arXiv

Interpretability

Explainable AI

Paper TitleYearConference/Journal
Debiasing synthetic data generated by deep generative models2024NeurIPS
Knowledge-driven AI-generated data for accurate and interpretable breast ultrasound diagnoses2024arXiv
Don't trust your eyes: on the (un) reliability of feature visualizations2023arXiv
Multifaceted feature visualization: Uncovering the different types of features learned by each neuron in deep neural networks2016arXiv
Understanding deep image representations by inverting them2015CVPR

Governance

Data watermarking

Paper TitleYearConference/Journal
Can watermarking large language models prevent copyrighted text generation and hide training data?2025AAAI
TimeWak: Temporal Chained-Hashing Watermark for Time Series Data2025arXiv

2.4 Embodied AI

Perception

Visual sensing

Paper TitleYearConference/Journal
Re$^3$Sim: Generating High-Fidelity Simulation Data via 3D-Photorealistic Real-to-Sim for Robotic Manipulation2025arXiv
Splatsim: Zero-shot sim2real transfer of rgb manipulation policies using gaussian splatting2025ICRA
Habitat 3.0: A Co-Habitat for Humans, Avatars, and Robots2024ICLR
Habitat synthetic scenes dataset (hssd-200): An analysis of 3d scene scale and realism tradeoffs for objectgoal navigation2024CVPR
Dextreme: Transfer of agile in-hand manipulation from simulation to reality2023ICRA
Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments2023RAL
ProcTHOR: Large-Scale Embodied AI Using Procedural Generation2022NeurIPS
MuJoCo: A physics engine for model-based control2012IROS

Force sensing

Paper TitleYearConference/Journal
Humanoid-Gym: Reinforcement Learning for Humanoid Robot with Zero-Shot Sim2Real Transfer2024arXiv
ARNOLD: A Benchmark for Language-Grounded Task Learning With Continuous States in Realistic 3D Scenes2023ICCV
Dextreme: Transfer of agile in-hand manipulation from simulation to reality2023ICRA

Sensor fusion

Paper TitleYearConference/Journal
SpatialVLM: Endowing Vision-Language Models with Spatial Reasoning Capabilities2024CVPR
Embodiedgpt: Vision-language pre-training via embodied chain of thought2023NeurIPS
PaLM-E: An Embodied Multimodal Language Model2023arXiv
Rt-2: Vision-language-action models transfer web knowledge to robotic control2023CoRL

Interaction

Trajectory synthesis

Paper TitleYearConference/Journal
Dexmimicgen: Automated data generation for bimanual dexterous manipulation via imitation learning2025ICRA
ReBot: Scaling Robot Learning with Real-to-Sim-to-Real Robotic Video Synthesis2025arXiv
3D Diffusion Policy: Generalizable Visuomotor Policy Learning via Simple 3D Representations2024RSS
Intervengen: Interventional data generation for robust and data-efficient robot imitation learning2024IROS
Diffusion Policy: Visuomotor Policy Learning via Action Diffusion2023RSS
MimicGen: A Data Generation System for Scalable Robot Learning using Human Demonstrations2023CoRL
Scaling robot learning with semantically imagined experience2023arXiv

Environment synthesis

Paper TitleYearConference/Journal
Holodeck: Language guided generation of 3d embodied ai environments2024CVPR
Partnr: A benchmark for planning and reasoning in embodied multi-agent tasks2024arXiv
RoboGen: Towards Unleashing Infinite Data for Automated Robot Learning via Generative Simulation2024ICML
Teach: Task-driven embodied agents that chat2022AAAI
Luminous: Indoor scene generation for embodied ai challenges2021arXiv
Alfred: A benchmark for interpreting grounded instructions for everyday tasks2020CVPR
Virtualhome: Simulating household activities via programs2018CVPR

Human behavior synthesis

Paper TitleYearConference/Journal
3D Human Reconstruction in the Wild with Synthetic Data Using Generative Models2024arXiv
Social-llava: Enhancing robot navigation through human-language reasoning in social spaces2024arXiv
Toward human-like social robot navigation: A large-scale, multi-modal, social human navigation dataset2023IROS
Socially compliant navigation dataset (scand): A large-scale dataset of demonstrations for social navigation2022RAL

Generalization

Cross-embodiment training

Paper TitleYearConference/Journal
Droid: A large-scale in-the-wild robot manipulation dataset2024arXiv
Octo: An open-source generalist robot policy2024arXiv
Open x-embodiment: Robotic learning datasets and rt-x models: Open x-embodiment collaboration 02024ICRA
Openvla: An open-source vision-language-action model2024arXiv

Vision-language-action models

Paper TitleYearConference/Journal
Diffusion forcing: Next-token prediction meets full-sequence diffusion2024NeurIPS
Learning universal policies via text-guided video generation2023NeurIPS
PaLM-E: An Embodied Multimodal Language Model2023arXiv
Rt-2: Vision-language-action models transfer web knowledge to robotic control2023CoRL

Sim-to-real transfer

Paper TitleYearConference/Journal
Re$^3$Sim: Generating High-Fidelity Simulation Data via 3D-Photorealistic Real-to-Sim for Robotic Manipulation2025arXiv
Splatsim: Zero-shot sim2real transfer of rgb manipulation policies using gaussian splatting2025ICRA
Transic: Sim-to-real policy transfer by learning from online correction2024arXiv
Dextreme: Transfer of agile in-hand manipulation from simulation to reality2023ICRA
Bi-directional domain adaptation for sim2real transfer of embodied navigation agents2021RAL
RetinaGAN: An Object-aware Approach to Sim-to-Real Transfer2020arXiv

3. Challenges & Future Directions

Model Collapse

Paper TitleYearConference/Journal
A theoretical perspective: How to prevent model collapse in self-consuming training loops2025arXiv
AI models collapse when trained on recursively generated data2024Nature
Beyond model collapse: Scaling up with synthesized data requires verification2024arXiv
Fairness feedback loops: training on synthetic data amplifies bias2024FACCT
Is model collapse inevitable? breaking the curse of recursion by accumulating real and synthetic data2024arXiv
Self-correcting self-consuming loops for generative model training2024arXiv
Towards theoretical understandings of self-consuming generative models2024arXiv
Large language models suffer from their own output: An analysis of the self-consuming training loop2023arXiv
On the stability of iterative retraining of generative models on their own data2023arXiv
Self-consuming generative models go mad2023ICLR

Utility-Privacy Tradeoffs

Generation-Evaluation Bias

Active Data Synthesis

Paper TitleYearConference/Journal
Controlled training data generation with diffusion models2024arXiv
Llm see, llm do: Guiding data generation to target non-differentiable objectives2024arXiv

Synthetic Data Evaluation

Paper TitleYearConference/Journal
A multi-faceted evaluation framework for assessing synthetic data generated by large language models2024arXiv

Multi-Modal Data Synthesis

Paper TitleYearConference/Journal
Enhanced Visual Instruction Tuning with Synthesized Image-Dialogue Data2024ACL
Strengthening multimodal large language model with bootstrapped preference optimization2024ECCV
Synthvlm: High-efficiency and high-quality synthetic data for vision language models2024CoRL

↑ Back to Top ↑

Contributors

WalkerWorldPeace

12 commits

Egg-Hu

6 commits

Similar repos

ahmad-alismail/LLM_based_Synthetic_Data_Generation

A curated and continuously updated collection of papers, tools, and datasets on synthetic data…

16

pengr/LLM-Synthetic-Data

A live reading list for LLM data synthesis (Updated to July, 2025).

492

wasiahmad/Awesome-LLM-Synthetic-Data

A reading list on LLM based Synthetic Data Generation 🔥

1,553

Zhen-Tan-dmml/LLM4Annotation

648

raojay7/Awesome-LLMs-Data-AI

A curated and automatically updated collection of LLM data research, covering data substrates, data…

109

Tebmer/Awesome-Knowledge-Distillation-of-LLMs

This repository collects papers for "A Survey on Knowledge Distillation of Large Language Models".…

1,314

daochenzha/data-centric-AI

A curated, but incomplete, list of data-centric AI resources.

1,157

susumuota/synthetic-data-hands-on

16

Egg-Hu/Awesome-Synthetic-Data-Generation

23

18 commits

updated Jan 7, 2026

See the code

README

Awesome Synthetic Data Generation

PRs Welcome Stars Forks

A comprehensive survey and curated collection of resources on synthetic data generation.

📚 Table of Contents (Click to Expand)

1. Methodologies

1.1 Generation-Based Synthesis

Synthesis from scratch

Paper TitleYearConference/Journal
Absolute zero: Reinforced self-play reasoning with zero data2025arXiv
Synthesizing post-training data for llms through multi-agent simulation2024arXiv
Learning vision from models rivals learning vision from data2024CVPR
Mathvista: Evaluating mathematical reasoning of foundation models in visual contexts2023arXiv
Dreamteacher: Pretraining image backbones with deep generative models2023ICCV
Self-instruct: Aligning language models with self-generated instructions2022arXiv
Training verifiers to solve math word problems2021arXiv

Synthesis from seeds

Paper TitleYearConference/Journal
Spread Preference Annotation: Direct Preference Judgment for Efficient LLM Alignment2025ICLR
Human-guided image generation for expanding small-scale training image datasets2025TVCG
Cream: Consistency regularized self-rewarding language models2024arXiv
Distribution-aware data expansion with diffusion models2024NeurIPS
Mitigating catastrophic forgetting in large language models with self-synthesized rehearsal2024arXiv
Towards automating text annotation: A case study on semantic proximity annotation using gpt-42024arXiv
Self-Rewarding Language Models2024arXiv
Real-fake: Effective training data synthesis through distribution matching2023arXiv
Expanding small-scale datasets with guided imagination2023NeurIPS
Training on thin air: Improve image classification with generated data2023arXiv

Synthesis from structure

Paper TitleYearConference/Journal
Synthesize-on-Graph: Knowledgeable Synthetic Data Generation for Continue Pre-training of Large Language Models2025arXiv
Logictree: Improving complex reasoning of LLMs via instantiated multi-step synthetic logical data2025NeurIPS
A Graph-Based Synthetic Data Pipeline for Scaling High-Quality Reasoning Instructions2024arXiv
MathScale: Scaling Instruction Tuning for Mathematical Reasoning2024ICML
ControlMath: Controllable Data Generation Promotes Math Generalist Models2024EMNLP
Enhancing logical reasoning in large language models through graph-based synthetic data2024arXiv
Infinitymath: A scalable instruction tuning dataset in programmatic mathematical reasoning2024CIKM
Synthetic data (almost) from scratch: Generalized instruction tuning for language models2024arXiv

Synthesis with evolution

Paper TitleYearConference/Journal
R-Zero: Self-Evolving Reasoning LLM from Zero Data2025arXiv
Prismatic Synthesis: Gradient-based Data Diversification Boosts Generalization in LLM Reasoning2025arXiv
HS-STAR: Hierarchical Sampling for Self-Taught Reasoners via Difficulty Estimation and Budget Reallocation2025arXiv
Enhancing LLM Reasoning with Iterative DPO: A Comprehensive Empirical Investigation2025arXiv
Spend wisely: Maximizing post-training gains in iterative synthetic data boostrapping2025arXiv
Scaling laws of synthetic images for model training... for now2024CVPR
Cream: Consistency regularized self-rewarding language models2024arXiv
Dualformer: Controllable fast and slow thinking by learning with randomized reasoning traces2024arXiv
Improving clip training with language rewrites2023NeurIPS
Feedback-guided data synthesis for imbalanced classification2023arXiv
Raft: Reward ranked finetuning for generative foundation model alignment2023arXiv
Synthetic data from diffusion models improves imagenet classification2023arXiv
Self-Rewarding Language Models2024arXiv
Spread Preference Annotation: Direct Preference Judgment for Efficient LLM Alignment2025ICLR
Star: Bootstrapping reasoning with reasoning2022NeurIPS
Unleashing reasoning capability of llms via scalable question synthesis from scratch2024arXiv
A Survey on LLM-as-a-Judge2024arXiv

1.2 Inversion-Based Synthesis

Data-space inversion

Paper TitleYearConference/Journal
Reverse-Engineered Reasoning for Open-Ended Generation2025arXiv
Dreaming to distill: Data-free knowledge transfer via deepinversion2020CVPR

Latent-space inversion

Paper TitleYearConference/Journal
Generative Model Inversion Through the Lens of the Manifold Hypothesis2025arXiv
Pseudo-private data guided model inversion attacks2024NeurIPS
Open-vocabulary customization from clip via data-free knowledge distillation2025ICLR
Reinforcement learning-based black-box model inversion attacks2023CVPR
Re-thinking model inversion attacks against deep neural networks2023CVPR
Learning to Learn from APIs: Black-Box Data-Free Meta-Learning2023arXiv
Label-only model inversion attacks via knowledge transfer2023NeurIPS
Label-only model inversion attacks via boundary repulsion2022CVPR
Mirror: Model inversion for deep learning network with high fidelity2022NDSS
Variational model inversion attacks2021NeurIPS
Knowledge-enriched distributional model inversion attacks2021ICCV
Contrastive model inversion for data-free knowledge distillation2021arXiv
Deep image prior2018CVPR
The secret revealer: Generative model-inversion attacks against deep neural networks2020CVPR
Plug & play attacks: Towards robust and flexible model inversion attacks2022arXiv

1.3 Simulation-Based Synthesis

Agent-based simulation

Paper TitleYearConference/Journal
Synthesizing post-training data for llms through multi-agent simulation2024arXiv
Autogen: Enabling next-gen LLM applications via multi-agent conversations2024COLM
Camel: Communicative agents for" mind" exploration of large language model society2023NeurIPS
MetaGPT: Meta programming for a multi-agent collaborative framework2023ICLR

Platform-based simulation

Paper TitleYearConference/Journal
Synthesizing post-training data for llms through multi-agent simulation2024arXiv
Autogen: Enabling next-gen LLM applications via multi-agent conversations2024COLM
Camel: Communicative agents for" mind" exploration of large language model society2023NeurIPS
MetaGPT: Meta programming for a multi-agent collaborative framework2023ICLR
Unity Perception: Generate Synthetic Data for Computer Vision2022arXiv
Omniverse Replicator: Synthetic Data Generation for AI2021arXiv

1.4 Augmentation-Based Synthesis

Rule-based augmentation

Paper TitleYearConference/Journal
Cutmix: Regularization strategy to train strong classifiers with localizable features2019ICCV
EDA: Easy data augmentation techniques for boosting performance on text classification tasks2019arXiv
mixup: Beyond empirical risk minimization2017arXiv

Generative augmentation

Paper TitleYearConference/Journal
Data augmentation for image classification using generative ai2025WACV
Self-improving diffusion models with synthetic data2024arXiv
On the diversity of synthetic data and its impact on training large language models2024arXiv
Scaling synthetic data creation with 1,000,000,000 personas2024arXiv
Synthetic continued pretraining2024arXiv
Dart-math: Difficulty-aware rejection tuning for mathematical problem-solving2024NeurIPS
Detdiffusion: Synergizing generative and perceptive models for enhanced data generation and perception2024CVPR
Diffusemix: Label-preserving data augmentation with diffusion models2024CVPR
Dreamda: Generative data augmentation with diffusion models2024arXiv
Learning to augment synthetic images for sim2real policy transfer2019IROS

2. Applications

2.1 Data-centric AI

Data Accessibility

Zero/Few-shot learning

Paper TitleYearConference/Journal
Absolute zero: Reinforced self-play reasoning with zero data2025arXiv
Generating synthetic datasets for few-shot prompt tuning2024arXiv
View-invariant policy learning via zero-shot novel view synthesis2024arXiv
Prompting-based synthetic data generation for few-shot question answering2024arXiv
Datadream: Few-shot guided dataset generation2024ECCV
Learning vision from models rivals learning vision from data2024CVPR
Tuning language models as training data generators for augmentation-enhanced few-shot learning2023PMLR

Federated learning

Paper TitleYearConference/Journal
Federated Knowledge Recycling: Privacy-preserving synthetic data sharing2025PATTERN RECOGN LETT
Synthetic data aided federated learning using foundation models2024arXiv
Understanding Data Reconstruction Leakage in Federated Learning from a Theoretical Perspective2024arXiv
Gradient inversion of federated diffusion models2024arXiv
VertiMRF: Differentially Private Vertical Federated Data Synthesis2024KDD
Data-Free Federated Class Incremental Learning with Diffusion-Based Generative Memory2024arXiv
Fine-tuning global model via data-free knowledge distillation for non-iid federated learning2022CVPR
DENSE: Data-Free One-Shot Federated Learning2022NeurIPS
Parameterized Data-free knowledge distillation for heterogeneous federated learning2021ICML

Data-free knowledge distillation

Paper TitleYearConference/Journal
Open-vocabulary customization from clip via data-free knowledge distillation2025ICLR
Synthetic image learning: Preserving performance and preventing membership inference attacks2025PATTERN RECOGN LETT
Aligning teacher with student preferences for tailored training data generation2024arXiv
Model Conversion via Differentially Private Data-Free Distillation2023arXiv
Data-Free Knowledge Distillation via Feature Exchange and Activation Region Constraint2023CVPR
Re-thinking model inversion attacks against deep neural networks2023CVPR
Up to 100x faster data-free knowledge distillation2022AAAI
Momentum Adversarial Distillation: Handling Large Distribution Shifts in Data-Free Knowledge Distillation2022NeurIPS
Dreaming to distill: Data-free knowledge transfer via deepinversion2020CVPR
Data-free knowledge distillation with soft targeted transfer set synthesis2021AAAI
Contrastive model inversion for data-free knowledge distillation2021arXiv
Data-free knowledge distillation for object detection2021WACV
Data-free learning of student networks2019ICCV

Data-free pruning/quantization

Paper TitleYearConference/Journal
Sharpness-aware data generation for zero-shot quantization2024arXiv
Distilled Pruning: Using Synthetic Data to Win the Lottery2023arXiv

Data-free meta-learning

Paper TitleYearConference/Journal
Task-distributionally robust data-free meta-learning2025TPAMI
FREE: Faster and Better Data-Free Meta-Learning2024CVPR
Meta-learning without data via unconditional diffusion models2024TCSVT
Task Groupings Regularization: Data-Free Meta-Learning with Heterogeneous Pre-trained Models2024arXiv
Architecture, Dataset and Model-Scale Agnostic Data-free Meta-Learning2023CVPR
Learning to Learn from APIs: Black-Box Data-Free Meta-Learning2023arXiv

Data-free continual learning

Paper TitleYearConference/Journal
Data-Free Federated Class Incremental Learning with Diffusion-Based Generative Memory2024arXiv
Mitigating catastrophic forgetting in large language models with self-synthesized rehearsal2024arXiv
Self-distillation bridges distribution gap in language model fine-tuning2024arXiv
A data-free approach to mitigate catastrophic forgetting in federated class incremental learning for vision tasks2023NeurIPS
Ddgr: Continual learning with deep diffusion-based generative replay2023ICML

Data Refinement

Dataset distillation

Paper TitleYearConference/Journal
Unlocking Dataset Distillation with Diffusion Models2025NeurIPS
Generative dataset distillation based on diffusion model2024ECCV
Dim: Distilling dataset into generative model2024ECCV
Dataset-distillation generative model for speech emotion recognition2024arXiv
Generalizing dataset distillation via deep generative prior2023CVPR
Dataset condensation with distribution matching2023PMLR
Dataset distillation using neural feature regression2022NeurIPS
Dataset distillation by matching training trajectories2022CVPR
Cafe: Learning to condense dataset by aligning features2022CVPR
Dataset condensation with gradient matching2020arXiv
Dataset distillation2018arXiv

Dataset purification

Paper TitleYearConference/Journal
Adbm: Adversarial diffusion bridge model for reliable adversarial purification2024arXiv
Better synthetic data by retrieving and transforming existing datasets2024arXiv
Dataelixir: Purifying poisoned dataset to mitigate backdoor attacks via diffusion models2024AAAI
Diffusion models for adversarial purification2022arXiv

2.2 Model-centric AI

General Model Enhancement

General ability

Paper TitleYearConference/Journal
Synthesize-on-Graph: Knowledgeable Synthetic Data Generation for Continue Pre-training of Large Language Models2025arXiv
Large language models, physics-based modeling, experimental measurements: the trinity of data-scarce learning of polymer properties2024arXiv
Nemotron-CC: Transforming Common Crawl into a refined long-horizon pretraining dataset2024arXiv
Rephrasing the web: A recipe for compute and data-efficient language modeling2024arXiv
Scilitllm: How to adapt llms for scientific literature understanding2024arXiv
Synthetic continued pretraining2024arXiv
VILA $^ 2$: VILA Augmented VILA2024arXiv
Fine-tuning or retrieval? comparing knowledge injection in llms2023arXiv
Llemma: An open language model for mathematics2023arXiv
Textbooks are all you need2023arXiv
Textbooks are all you need ii: phi-1.5 technical report2023arXiv
Tinystories: How small can language models be and still speak coherent english?2023arXiv
Enhancing multilingual language model with massive multilingual knowledge triples2021arXiv

Domain Model Enhancement

Reasoning

Paper TitleYearConference/Journal
Absolute zero: Reinforced self-play reasoning with zero data2025arXiv
HS-STAR: Hierarchical Sampling for Self-Taught Reasoners via Difficulty Estimation and Budget Reallocation2025arXiv
Learning to Pose Problems: Reasoning-Driven and Solver-Adaptive Data Synthesis for Large Reasoning Models2025arXiv
Logictree: Improving complex reasoning of LLMs via instantiated multi-step synthetic logical data2025NeurIPS
Prismatic Synthesis: Gradient-based Data Diversification Boosts Generalization in LLM Reasoning2025arXiv
Seed-Coder: Let the Code Model Curate Data for Itself2025arXiv
Spread Preference Annotation: Direct Preference Judgment for Efficient LLM Alignment2025ICLR
Synthesize-on-Graph: Knowledgeable Synthetic Data Generation for Continue Pre-training of Large Language Models2025arXiv
Thinking LLMs: General Instruction Following with Thought Generation2025ICML
Unleashing Reasoning Capability of LLMs via Scalable Question Synthesis from Scratch2025arXiv
A Graph-Based Synthetic Data Pipeline for Scaling High-Quality Reasoning Instructions2024arXiv
Aligning teacher with student preferences for tailored training data generation2024arXiv
Autocoder: Enhancing code large language model with$\backslash$textsc $$AIEV-Instruct$$2024arXiv
Boosting reward model with preference-conditional multi-aspect synthetic data generation2024arXiv
ControlMath: Controllable Data Generation Promotes Math Generalist Models2024EMNLP
From the Least to the Most: Building a Plug-and-Play Visual Reasoner via Data Synthesis2024EMNLP
HexaCoder: Secure Code Generation via Oracle-Guided Synthetic Training Data2024arXiv
Infinitymath: A scalable instruction tuning dataset in programmatic mathematical reasoning2024CIKM
Jiuzhang3. 0: Efficiently improving mathematical reasoning by training small data synthesis models2024NeurIPS
MAmmoTH: Building Math Generalist Models through Hybrid Instruction Tuning2024ICLR
Marco-o1: Towards open reasoning models for open-ended solutions2024arXiv
MathScale: Scaling Instruction Tuning for Mathematical Reasoning2024ICML
MetaMath: Bootstrap Your Own Mathematical Questions for Large Language Models2024ICLR
Openmathinstruct-1: A 1.8 million math instruction tuning dataset2024NeurIPS
Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking2024COLM
Refined direct preference optimization with synthetic data for behavioral alignment of llms2024arXiv
Rl on incorrect synthetic data scales the efficiency of llm math reasoning by eight-fold2024NeurIPS
Self-Consistency Preference Optimization2024arXiv
Self-play with execution feedback: Improving instruction-following capabilities of large language models2024arXiv
Self-Rewarding Language Models2024arXiv
Small Language Models Need Strong Verifiers to Self-Correct Reasoning2024ACL
Strengthening multimodal large language model with bootstrapped preference optimization2024ECCV
ToRA: A Tool-Integrated Reasoning Agent for Mathematical Problem Solving2024ICLR
Tree-instruct: A preliminary study of the intrinsic relationship between complexity and alignment2024COLING
WizardLM: Empowering large pre-trained language models to follow complex instructions2024ICLR
Lamini-lm: A diverse herd of distilled models from large-scale instructions2023arXiv
Mathcoder: Seamless code integration in llms for enhanced mathematical reasoning2023arXiv
Reflection-tuning: Recycling data for better instruction-tuning2023NeurIPS
What makes good data for alignment? a comprehensive study of automatic data selection in instruction tuning2023arXiv
Wizardcoder: Empowering code large language models with evol-instruct2023arXiv
Wizardmath: Empowering mathematical reasoning for large language models via reinforced evol-instruct2023arXiv
Star: Bootstrapping reasoning with reasoning2022NeurIPS

Code

Paper TitleYearConference/Journal
Absolute zero: Reinforced self-play reasoning with zero data2025arXiv
Case2Code: Scalable Synthetic Data for Code Generation2025COLING
Increasing LLM Coding Capabilities through Diverse Synthetic Coding Tasks2025arXiv
Seed-Coder: Let the Code Model Curate Data for Itself2025arXiv
Autocoder: Enhancing code large language model with$\backslash$textsc $$AIEV-Instruct$$2024arXiv
Codeclm: Aligning language models with tailored synthetic data2024arXiv
HexaCoder: Secure Code Generation via Oracle-Guided Synthetic Training Data2024arXiv
WizardLM: Empowering large pre-trained language models to follow complex instructions2024ICLR
Wizardcoder: Empowering code large language models with evol-instruct2023arXiv

Instruction following

Paper TitleYearConference/Journal
Recursive introspection: Teaching language model agents how to self-improve2024NeurIPS
Selective reflection-tuning: Student-selected data recycling for llm instruction-tuning2024ACL
Self-play with execution feedback: Improving instruction-following capabilities of large language models2024arXiv
Self-refine instruction-tuning for aligning reasoning in language models2024arXiv
Spar: Self-play with tree-search refinement to improve instruction-following in large language models2024arXiv
Tree-instruct: A preliminary study of the intrinsic relationship between complexity and alignment2024LREC-COLING
WizardLM: Empowering large pre-trained language models to follow complex instructions2024ICLR
# instag: Instruction tagging for analyzing supervised fine-tuning of large language models2023arXiv
Alpagasus: Training a better alpaca with fewer data2023arXiv
Lamini-lm: A diverse herd of distilled models from large-scale instructions2023arXiv
Lima: Less is more for alignment2023NeurIPS
Reflection-tuning: Recycling data for better instruction-tuning2023NeurIPS
What makes good data for alignment? a comprehensive study of automatic data selection in instruction tuning2023arXiv
Self-instruct: Aligning language models with self-generated instructions2022arXiv

Preference

Paper TitleYearConference/Journal
Spread Preference Annotation: Direct Preference Judgment for Efficient LLM Alignment2025ICLR
Aligning teacher with student preferences for tailored training data generation2024arXiv
Boosting reward model with preference-conditional multi-aspect synthetic data generation2024arXiv
Course-correction: Safety alignment using synthetic preferences2024arXiv
Refined direct preference optimization with synthetic data for behavioral alignment of llms2024arXiv
Self-Consistency Preference Optimization2024arXiv
Self-directed synthetic dialogues and revisions technical report2024arXiv
Strengthening multimodal large language model with bootstrapped preference optimization2024ECCV

In-context learning

Paper TitleYearConference/Journal
Embracing Collaboration Over Competition: Condensing Multiple Prompts for Visual In-Context Learning2025CVPR
Self-Harmonized Chain of Thought2025arXiv
Auto-ICL: In-Context Learning without Human Supervision2024arXiv
Demonstration Augmentation for Zero-shot In-context Learning2024ACL
Self-Prompting Large Language Models for Zero-Shot Open-Domain QA2024arXiv
Automatic Chain of Thought Prompting in Large Language Models2023ICLR
Generate Rather than Retrieve: Large Language Models Are Strong Context Generators2023arXiv
Privacy-preserving in-context learning with differentially private few-shot generation2023arXiv
Recitation-Augmented Language Models2023arXiv
Self-ICL: Zero-Shot In-Context Learning with Self-Generated Demonstrations2023EMNLP
Self-Generated In-Context Learning: Leveraging Auto-regressive Language Models as a Demonstration Generator2022arXiv

Reinforcement Learning

Paper TitleYearConference/Journal
Kimi K2: Open Agentic Intelligence2025arXiv
s1: Simple test-time scaling2025arXiv
Synthetic Data Generation & Multi-Step RL for Reasoning & Tool Use2025arXiv
Synthetic Data RL: Task Definition Is All You Need2025arXiv
RL on Incorrect Synthetic Data Scales the Efficiency of LLM Math Reasoning by Eight-Fold2024arXiv

Model Evaluation

Synthetic benchmark

Paper TitleYearConference/Journal
Achilles' Heel of Mamba: Essential difficulties of the Mamba architecture demonstrated by synthetic data2025arXiv
Measuring General Intelligence with Generated Games2025arXiv
Efficacy of Synthetic Data as a Benchmark2024arXiv
Can you rely on your model evaluation? improving model evaluation with synthetic test data2023NeurIPS
SynBench: Task-Agnostic Benchmarking of Pretrained Representations using Synthetic Data2022NeurIPS

2.3 Trustworthy AI

Privacy

Privacy-preserving learning

Paper TitleYearConference/Journal
Does Training with Synthetic Data Truly Protect Privacy?2025arXiv
Federated Knowledge Recycling: Privacy-preserving synthetic data sharing2025PATTERN RECOGN LETT
Synthetic image learning: Preserving performance and preventing membership inference attacks2025PATTERN RECOGN LETT
Differentially Private Synthetic Data via Foundation Model APIs 1: Images)2024ICLR
Differentially private synthetic data via foundation model APIs 2: Text2024arXiv
Enhancing the utility of privacy-preserving cancer classification using synthetic data2024MICCAI
Generate synthetic text approximating the private distribution with differential privacy2024IJCAI
Privacy-preserving instructions for aligning large language models2024arXiv
SETSUBUN: Revisiting Membership Inference Game for Evaluating Synthetic Data Generation2024J INF PROCESS SYST
Synthetic data aided federated learning using foundation models2024arXiv
Towards Privacy-Preserving Relational Data Synthesis via Probabilistic Relational Models2024German Conference on Artificial Intellig
Using synthetic data to mitigate unfairness and preserve privacy through single-shot federated learning2024arXiv
VertiMRF: Differentially Private Vertical Federated Data Synthesis2024KDD

Model inversion attack

Paper TitleYearConference/Journal
Gradient inversion of federated diffusion models2024arXiv
Understanding Data Reconstruction Leakage in Federated Learning from a Theoretical Perspective2024arXiv
The secret revealer: Generative model-inversion attacks against deep neural networks2020CVPR
A methodology for formalizing model-inversion attacks2016CSF
Model inversion attacks that exploit confidence information and basic countermeasures2015SIGSAC

Safety & Security

Model stealing attack

Paper TitleYearConference/Journal
Unifying Multimodal Large Language Model Capabilities and Modalities via Model Merging2025arXiv
Data-free model extraction2021CVPR
Maze: Data-free model stealing attack using zeroth-order gradient estimation2021CVPR

Adversarial defense

Paper TitleYearConference/Journal
Deceptive diffusion: generating synthetic adversarial examples2025SSVM
Is synthetic data all we need? benchmarking the robustness of models trained with synthetic images2024CVPR
TSynD: Targeted Synthetic Data Generation for Enhanced Medical Image Classification: Leveraging Epistemic Uncertainty to Improve Model Performance2024arXiv
Leaving reality to imagination: Robust classification via generated datasets2023arXiv
Robust learning meets generative models: Can proxy distributions improve adversarial robustness?2021arXiv
Explaining and harnessing adversarial examples2014arXiv

Machine unlearning

Paper TitleYearConference/Journal
Mitigating catastrophic forgetting in large language models with self-synthesized rehearsal2024arXiv

Fairness

De-bias learning

Paper TitleYearConference/Journal
Generating Informative Samples for Risk-Averse Fine-Tuning of Downstream Tasks2025NeurIPS
Know" No''Better: A Data-Driven Approach for Enhancing Negation Awareness in CLIP2025arXiv
Modeling multi-task model merging as adaptive projective gradient descent2025ICML
Virus Infection Attack on LLMs: Your Poisoning Can Spread" VIA" Synthetic Data2025arXiv
FADE: Towards Fairness-aware Generation for Domain Generalization via Classifier-Guided Score-based Diffusion Models2024arXiv
Refined direct preference optimization with synthetic data for behavioral alignment of llms2024ICML
Strengthening multimodal large language model with bootstrapped preference optimization2024ECCV
SynthFair: Ensuring Subgroup Fairness in Classification via Synthetic Data Generation2024CSCE
Using synthetic data to mitigate unfairness and preserve privacy through single-shot federated learning2024arXiv
Discover and cure: Concept-aware mitigation of spurious correlation2023ICML

Long-tail learning

Paper TitleYearConference/Journal
DALK: Dynamic Co-Augmentation of LLMs and KG to answer Alzheimer's Disease Questions with Scientific Literature2024arXiv
Ltgc: Long-tail recognition via leveraging llms-driven generated content2024CVPR
Feedback-guided data synthesis for imbalanced classification2023arXiv

Interpretability

Explainable AI

Paper TitleYearConference/Journal
Debiasing synthetic data generated by deep generative models2024NeurIPS
Knowledge-driven AI-generated data for accurate and interpretable breast ultrasound diagnoses2024arXiv
Don't trust your eyes: on the (un) reliability of feature visualizations2023arXiv
Multifaceted feature visualization: Uncovering the different types of features learned by each neuron in deep neural networks2016arXiv
Understanding deep image representations by inverting them2015CVPR

Governance

Data watermarking

Paper TitleYearConference/Journal
Can watermarking large language models prevent copyrighted text generation and hide training data?2025AAAI
TimeWak: Temporal Chained-Hashing Watermark for Time Series Data2025arXiv

2.4 Embodied AI

Perception

Visual sensing

Paper TitleYearConference/Journal
Re$^3$Sim: Generating High-Fidelity Simulation Data via 3D-Photorealistic Real-to-Sim for Robotic Manipulation2025arXiv
Splatsim: Zero-shot sim2real transfer of rgb manipulation policies using gaussian splatting2025ICRA
Habitat 3.0: A Co-Habitat for Humans, Avatars, and Robots2024ICLR
Habitat synthetic scenes dataset (hssd-200): An analysis of 3d scene scale and realism tradeoffs for objectgoal navigation2024CVPR
Dextreme: Transfer of agile in-hand manipulation from simulation to reality2023ICRA
Orbit: A Unified Simulation Framework for Interactive Robot Learning Environments2023RAL
ProcTHOR: Large-Scale Embodied AI Using Procedural Generation2022NeurIPS
MuJoCo: A physics engine for model-based control2012IROS

Force sensing

Paper TitleYearConference/Journal
Humanoid-Gym: Reinforcement Learning for Humanoid Robot with Zero-Shot Sim2Real Transfer2024arXiv
ARNOLD: A Benchmark for Language-Grounded Task Learning With Continuous States in Realistic 3D Scenes2023ICCV
Dextreme: Transfer of agile in-hand manipulation from simulation to reality2023ICRA

Sensor fusion

Paper TitleYearConference/Journal
SpatialVLM: Endowing Vision-Language Models with Spatial Reasoning Capabilities2024CVPR
Embodiedgpt: Vision-language pre-training via embodied chain of thought2023NeurIPS
PaLM-E: An Embodied Multimodal Language Model2023arXiv
Rt-2: Vision-language-action models transfer web knowledge to robotic control2023CoRL

Interaction

Trajectory synthesis

Paper TitleYearConference/Journal
Dexmimicgen: Automated data generation for bimanual dexterous manipulation via imitation learning2025ICRA
ReBot: Scaling Robot Learning with Real-to-Sim-to-Real Robotic Video Synthesis2025arXiv
3D Diffusion Policy: Generalizable Visuomotor Policy Learning via Simple 3D Representations2024RSS
Intervengen: Interventional data generation for robust and data-efficient robot imitation learning2024IROS
Diffusion Policy: Visuomotor Policy Learning via Action Diffusion2023RSS
MimicGen: A Data Generation System for Scalable Robot Learning using Human Demonstrations2023CoRL
Scaling robot learning with semantically imagined experience2023arXiv

Environment synthesis

Paper TitleYearConference/Journal
Holodeck: Language guided generation of 3d embodied ai environments2024CVPR
Partnr: A benchmark for planning and reasoning in embodied multi-agent tasks2024arXiv
RoboGen: Towards Unleashing Infinite Data for Automated Robot Learning via Generative Simulation2024ICML
Teach: Task-driven embodied agents that chat2022AAAI
Luminous: Indoor scene generation for embodied ai challenges2021arXiv
Alfred: A benchmark for interpreting grounded instructions for everyday tasks2020CVPR
Virtualhome: Simulating household activities via programs2018CVPR

Human behavior synthesis

Paper TitleYearConference/Journal
3D Human Reconstruction in the Wild with Synthetic Data Using Generative Models2024arXiv
Social-llava: Enhancing robot navigation through human-language reasoning in social spaces2024arXiv
Toward human-like social robot navigation: A large-scale, multi-modal, social human navigation dataset2023IROS
Socially compliant navigation dataset (scand): A large-scale dataset of demonstrations for social navigation2022RAL

Generalization

Cross-embodiment training

Paper TitleYearConference/Journal
Droid: A large-scale in-the-wild robot manipulation dataset2024arXiv
Octo: An open-source generalist robot policy2024arXiv
Open x-embodiment: Robotic learning datasets and rt-x models: Open x-embodiment collaboration 02024ICRA
Openvla: An open-source vision-language-action model2024arXiv

Vision-language-action models

Paper TitleYearConference/Journal
Diffusion forcing: Next-token prediction meets full-sequence diffusion2024NeurIPS
Learning universal policies via text-guided video generation2023NeurIPS
PaLM-E: An Embodied Multimodal Language Model2023arXiv
Rt-2: Vision-language-action models transfer web knowledge to robotic control2023CoRL

Sim-to-real transfer

Paper TitleYearConference/Journal
Re$^3$Sim: Generating High-Fidelity Simulation Data via 3D-Photorealistic Real-to-Sim for Robotic Manipulation2025arXiv
Splatsim: Zero-shot sim2real transfer of rgb manipulation policies using gaussian splatting2025ICRA
Transic: Sim-to-real policy transfer by learning from online correction2024arXiv
Dextreme: Transfer of agile in-hand manipulation from simulation to reality2023ICRA
Bi-directional domain adaptation for sim2real transfer of embodied navigation agents2021RAL
RetinaGAN: An Object-aware Approach to Sim-to-Real Transfer2020arXiv

3. Challenges & Future Directions

Model Collapse

Paper TitleYearConference/Journal
A theoretical perspective: How to prevent model collapse in self-consuming training loops2025arXiv
AI models collapse when trained on recursively generated data2024Nature
Beyond model collapse: Scaling up with synthesized data requires verification2024arXiv
Fairness feedback loops: training on synthetic data amplifies bias2024FACCT
Is model collapse inevitable? breaking the curse of recursion by accumulating real and synthetic data2024arXiv
Self-correcting self-consuming loops for generative model training2024arXiv
Towards theoretical understandings of self-consuming generative models2024arXiv
Large language models suffer from their own output: An analysis of the self-consuming training loop2023arXiv
On the stability of iterative retraining of generative models on their own data2023arXiv
Self-consuming generative models go mad2023ICLR

Utility-Privacy Tradeoffs

Generation-Evaluation Bias

Active Data Synthesis

Paper TitleYearConference/Journal
Controlled training data generation with diffusion models2024arXiv
Llm see, llm do: Guiding data generation to target non-differentiable objectives2024arXiv

Synthetic Data Evaluation

Paper TitleYearConference/Journal
A multi-faceted evaluation framework for assessing synthetic data generated by large language models2024arXiv

Multi-Modal Data Synthesis

Paper TitleYearConference/Journal
Enhanced Visual Instruction Tuning with Synthesized Image-Dialogue Data2024ACL
Strengthening multimodal large language model with bootstrapped preference optimization2024ECCV
Synthvlm: High-efficiency and high-quality synthetic data for vision language models2024CoRL

↑ Back to Top ↑

Similar repos

ahmad-alismail/LLM_based_Synthetic_Data_Generation

A curated and continuously updated collection of papers, tools, and datasets on synthetic data…

16

pengr/LLM-Synthetic-Data

A live reading list for LLM data synthesis (Updated to July, 2025).

492

wasiahmad/Awesome-LLM-Synthetic-Data

A reading list on LLM based Synthetic Data Generation 🔥

1,553

Zhen-Tan-dmml/LLM4Annotation

648

raojay7/Awesome-LLMs-Data-AI

A curated and automatically updated collection of LLM data research, covering data substrates, data…

109

Tebmer/Awesome-Knowledge-Distillation-of-LLMs

This repository collects papers for "A Survey on Knowledge Distillation of Large Language Models".…

1,314

daochenzha/data-centric-AI

A curated, but incomplete, list of data-centric AI resources.

1,157

susumuota/synthetic-data-hands-on

16

Contributors

WalkerWorldPeace

12 commits

Egg-Hu

6 commits