{"total":82,"rows":[{"id":"2404.01413","title":"Is Model Collapse Inevitable? Breaking the Curse of Recursion by Accumulating Real and Synthetic Data","created":"2024-04-01","cited_by_count":151,"area":"synth"},{"id":"2406.14532","title":"RL on Incorrect Synthetic Data Scales the Efficiency of LLM Math Reasoning by Eight-Fold","created":"2024-06-20","cited_by_count":99,"area":"lr"},{"id":"2404.05090","title":"How Bad is Training on Synthetic Data? A Statistical Analysis of Language Model Collapse","created":"2024-04-07","cited_by_count":79,"area":"synth"},{"id":"2410.04840","title":"Strong Model Collapse","created":"2024-10-07","cited_by_count":65,"area":"synth"},{"id":"2410.16713","title":"Collapse or Thrive? Perils and Promises of Synthetic Data in a Self-Generating World","created":"2024-10-22","cited_by_count":52,"area":null},{"id":"2410.09732","title":"LOKI: A Comprehensive Synthetic Data Detection Benchmark using Large Multimodal Models","created":"2024-10-13","cited_by_count":51,"area":"mllm"},{"id":"2406.07515","title":"Beyond Model Collapse: Scaling Up with Synthesized Data Requires Verification","created":"2024-06-11","cited_by_count":45,"area":"synth"},{"id":"2407.09499","title":"Self-Consuming Generative Models with Curated Data Provably Optimize Human Preferences","created":"2024-06-12","cited_by_count":39,"area":"alignment"},{"id":"2408.16333","title":"Self-Improving Diffusion Models with Synthetic Data","created":"2024-08-29","cited_by_count":35,"area":"diffusion"},{"id":"2503.03150","title":"Position: Model Collapse Does Not Mean What You Think","created":"2025-03-05","cited_by_count":33,"area":"synth"},{"id":"2410.22812","title":"Universality of the $\\pi^2/6$ Pathway in Avoiding Model Collapse","created":"2024-10-30","cited_by_count":26,"area":null},{"id":"2412.14689","title":"How to Synthesize Text Data without Model Collapse?","created":"2024-12-19","cited_by_count":25,"area":"synth"},{"id":"2412.17646","title":"Rate of Model Collapse in Recursive Training","created":"2024-12-23","cited_by_count":25,"area":"synth"},{"id":"2405.15699","title":"Dimension-free deterministic equivalents and scaling laws for random feature regression","created":"2024-05-24","cited_by_count":23,"area":null},{"id":"2412.02674","title":"Mind the Gap: Examining the Self-Improvement Capabilities of Large Language Models","created":"2024-12-03","cited_by_count":18,"area":"synth"},{"id":"2509.16499","title":"A Closer Look at Model Collapse: From a Generalization-to-Memorization Perspective","created":"2025-09-20","cited_by_count":18,"area":"diffusion"},{"id":"2502.18865","title":"A Theoretical Perspective: How to Prevent Model Collapse in Self-consuming Training Loops","created":"2025-02-26","cited_by_count":16,"area":"synth"},{"id":"2503.16322","title":"Ultra-Resolution Adaptation with Ease","created":"2025-03-20","cited_by_count":13,"area":"diffusion"},{"id":"2505.19046","title":"When Models Don't Collapse: On the Consistency of Iterative MLE","created":"2025-05-25","cited_by_count":13,"area":"synth"},{"id":"2407.17493","title":"Model Collapse in the Self-Consuming Chain of Diffusion Finetuning: A Novel Perspective from Quantitative Trait Modeling","created":"2024-07-04","cited_by_count":12,"area":"synth"},{"id":"2410.15234","title":"Bias Amplification: Large Language Models as Increasingly Biased Media","created":"2024-10-19","cited_by_count":12,"area":"synth"},{"id":"2502.08924","title":"Escaping Collapse: The Strength of Weak Data for Large Language Model Training","created":"2025-02-13","cited_by_count":11,"area":null},{"id":"2502.18049","title":"Recursive Learning Without Collapse: A Weighting-Based Stabilization Framework","created":"2025-02-25","cited_by_count":10,"area":"synth"},{"id":"2410.09005","title":"Analyzing Neural Scaling Laws in Two-Layer Networks with Power-Law Data Spectra","created":"2024-10-11","cited_by_count":9,"area":null},{"id":"2509.22341","title":"Preventing Model Collapse Under Overparametrization: Optimal Mixing Ratios for Interpolation Learning and Ridge Regression","created":"2025-09-26","cited_by_count":9,"area":"synth"},{"id":"2505.18346","title":"On the Mechanisms of Weak-to-Strong Generalization: A Theoretical Perspective","created":"2025-05-23","cited_by_count":8,"area":null},{"id":"2511.03492","title":"Why Less is More (Sometimes): A Theory of Data Curation","created":"2025-11-05","cited_by_count":8,"area":null},{"id":"2408.14960","title":"Multilingual Arbitrage: Optimizing Data Pools to Accelerate Multilingual Progress","created":"2024-08-27","cited_by_count":7,"area":null},{"id":"2410.08942","title":"Maximizing the Potential of Synthetic Data: Insights from Random Matrix Theory","created":"2024-10-11","cited_by_count":7,"area":"synth"},{"id":"2506.19882","title":"Position: Machine Learning Conferences Should Establish a \"Refutations and Critiques\" Track","created":"2025-06-24","cited_by_count":7,"area":null},{"id":"2407.01490","title":"LLM See, LLM Do: Guiding Data Generation to Target Non-Differentiable Objectives","created":"2024-07-01","cited_by_count":6,"area":null},{"id":"2505.21677","title":"What happens when generative AI models train recursively on each others' outputs?","created":"2025-05-27","cited_by_count":6,"area":null},{"id":"2510.16657","title":"Escaping Model Collapse via Synthetic Data Verification: Near-term Improvements and Long-term Convergence","created":"2025-10-18","cited_by_count":6,"area":"synth"},{"id":"2604.26965","title":"The Impact of AI-Generated Text on the Internet","created":"2026-04-14","cited_by_count":6,"area":null},{"id":"2603.11784","title":"Language Generation with Replay: A Learning-Theoretic View of Model Collapse","created":"2026-03-12","cited_by_count":5,"area":"synth"},{"id":"2412.14872","title":"Theoretical Proof that Auto-regressive Language Models Collapse when Real-world Data is a Finite Set","created":"2024-12-19","cited_by_count":4,"area":null},{"id":"2506.23174","title":"Data Can Speak for Itself: Quality-guided Utilization of Wireless Synthetic Data","created":"2025-06-29","cited_by_count":4,"area":null},{"id":"2602.16601","title":"Quantifying Error Propagation and Model Collapse in Diffusion Models","created":"2026-02-18","cited_by_count":4,"area":"diffusion"},{"id":"2510.17421","title":"Diffusion Models as Dataset Distillation Priors","created":"2025-10-20","cited_by_count":3,"area":"diffusion"},{"id":"2601.06160","title":"Student Guides Teacher: Weak-to-Strong Inference via Spectral Orthogonal Exploration","created":"2026-01-06","cited_by_count":3,"area":"lr"},{"id":"2605.14831","title":"Interestingness as an Inductive Heuristic for Future Compression Progress","created":"2026-05-14","cited_by_count":3,"area":null},{"id":"2410.12341","title":"Learning by Surprise: Adaptive Mitigation of Model Collapse in Large Language Models","created":"2024-10-16","cited_by_count":2,"area":"synth"},{"id":"2501.18962","title":"Spend Wisely: Maximizing Post-Training Gains in Iterative Synthetic Data Bootstrapping","created":"2025-01-31","cited_by_count":2,"area":"synth"},{"id":"2503.08117","title":"Convergence Dynamics and Stabilization Strategies of Co-Evolving Generative Models","created":"2025-03-11","cited_by_count":2,"area":"synth"},{"id":"2504.03814","title":"Recursive Training Loops in LLMs: How training data properties modulate distribution shift in generated data?","created":"2025-04-04","cited_by_count":2,"area":null},{"id":"2506.05101","title":"Privacy Amplification Through Synthetic Data: Insights from Linear Regression","created":"2025-06-05","cited_by_count":2,"area":null},{"id":"2510.08095","title":"Beyond Real Data: Synthetic Data through the Lens of Regularization","created":"2025-10-09","cited_by_count":2,"area":null},{"id":"2511.12804","title":"The Alignment Game: A Theory of Long-Horizon Alignment Through Recursive Curation","created":"2025-11-16","cited_by_count":2,"area":"alignment"},{"id":"2601.05184","title":"Observations and Remedies for Large Language Model Bias in Self-Consuming Performative Loop","created":"2026-01-08","cited_by_count":2,"area":null},{"id":"2607.04277","title":"Self-Reference in Large Language Models: The Introspection Threshold for Recursive Self-Improvement","created":"2026-07-05","cited_by_count":2,"area":"synth"},{"id":"2505.19550","title":"Turing Test 2.0: The General Intelligence Threshold","created":"2025-05-26","cited_by_count":1,"area":null},{"id":"2506.09401","title":"A theoretical basis for model collapse in recursive training","created":"2025-06-11","cited_by_count":1,"area":null},{"id":"2506.19262","title":"What Matters in LLM-generated Data: Diversity and Its Effect on Model Fine-Tuning","created":"2025-06-24","cited_by_count":1,"area":"synth"},{"id":"2509.09629","title":"Bridging the Capability Gap: Joint Alignment Tuning for Harmonizing LLM-based Multi-Agent Systems","created":"2025-09-11","cited_by_count":1,"area":"agents"},{"id":"2509.20101","title":"First-Extinction Law for Resampling Processes","created":"2025-09-24","cited_by_count":1,"area":null},{"id":"2511.13640","title":"Data Value in the Age of Scaling: Understanding LLM Scaling Dynamics Under Real-Synthetic Data Mixtures","created":"2025-11-17","cited_by_count":1,"area":"synth"},{"id":"2601.02563","title":"Compressed code: the hidden effects of quantization and distillation on programming tokens","created":"2026-01-05","cited_by_count":1,"area":null},{"id":"2601.22513","title":"Why Self-Rewarding Works: Theoretical Guarantees for Iterative Alignment of Language Models","created":"2026-01-30","cited_by_count":1,"area":"alignment"},{"id":"2602.10531","title":"From Collapse to Improvement: Statistical Perspectives on the Evolutionary Dynamics of Iterative Training on Contaminated Sources","created":"2026-02-11","cited_by_count":1,"area":"synth"},{"id":"2602.14029","title":"Why Self-Training Helps and Hurts: Denoising vs. Signal Forgetting","created":"2026-02-15","cited_by_count":1,"area":"distill"},{"id":"2605.04127","title":"Position: the Stochastic Parrot in the Coal Mine. Model Collapse is a Threat to Low-Resource Communities","created":"2026-05-05","cited_by_count":1,"area":"synth"},{"id":"2605.16245","title":"AI-Mediated Communication Can Steer Collective Opinion","created":"2026-05-15","cited_by_count":1,"area":null},{"id":"2605.23054","title":"Model Collapse as Cultural Evolution","created":"2026-05-21","cited_by_count":1,"area":"synth"},{"id":"2606.13732","title":"When Sample Selection Bias Precipitates Model Collapse","created":"2026-06-11","cited_by_count":1,"area":"synth"},{"id":"2607.17043","title":"Learning from Synthetic Data without Model Collapse in Iterative Instruction Tuning","created":"2026-07-19","cited_by_count":1,"area":"synth"},{"id":"2506.20623","title":"Lost in Retraining: Roaming the Parameter Space of Exponential Families Under Closed-Loop Learning","created":"2025-06-25","cited_by_count":0,"area":null},{"id":"2602.04895","title":"Privacy Amplification Persists under Unlimited Synthetic Data Release","created":"2026-02-03","cited_by_count":0,"area":null},{"id":"2602.16065","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","created":"2026-02-17","cited_by_count":0,"area":null},{"id":"2603.12683","title":"Experimental evidence of progressive ChatGPT models self-convergence","created":"2026-03-13","cited_by_count":0,"area":"synth"},{"id":"2605.19483","title":"Adynamical systems view of training generativemodels and the memorization phenomenon","created":"2026-05-19","cited_by_count":0,"area":null},{"id":"2606.02339","title":"Entropy Minimization without Model Collapse: Mitigating Prediction Bias in Medical Imaging","created":"2026-06-01","cited_by_count":0,"area":null},{"id":"2606.07382","title":"Covariance Shrinkage via Stochastic Interpolation","created":"2026-06-05","cited_by_count":0,"area":null},{"id":"2606.13796","title":"Recursively Trained Diffusion Models: Limiting Collapse Distribution and Spectral Characterization","created":"2026-06-11","cited_by_count":0,"area":"diffusion"},{"id":"2607.15450","title":"Prediction-Only Distillation in Linear and Logistic Regression","created":"2026-07-16","cited_by_count":0,"area":"distill"},{"id":"2608.21366","title":"Reviewing Model Collapse and Countermeasures","created":"2026-06-17","cited_by_count":0,"area":"synth"},{"id":"2608.31091","title":"Minimax bounds for watermarked and masked recursive discrete distribution estimation","created":"2026-08-31","cited_by_count":0,"area":null},{"id":"2609.07153","title":"FreqBLiMP: Frequency-Controlled Minimal Pairs Reveal Robustness and Fragility of LLMs Under Lexical Rarity","created":"2026-09-07","cited_by_count":0,"area":null},{"id":"2609.11149","title":"Break Step: Recursive Training Resonates with Replayed Sampling Noise","created":"2026-09-10","cited_by_count":0,"area":null},{"id":"2609.12353","title":"SynthSentry: Detecting Synthetic Data Contamination in Language Model Training Data","created":"2026-09-11","cited_by_count":0,"area":"synth"},{"id":"2610.01318","title":"Feature Selective Model Collapse in Diffusion Models: Total Replacement versus Fixed-Budget Training","created":"2026-10-01","cited_by_count":0,"area":"diffusion"},{"id":"2610.01674","title":"Invent a Dataset: Measuring dataset generation abilities with zero seed","created":"2026-10-01","cited_by_count":0,"area":null},{"id":"2610.15617","title":"Screening Collapse under Recursive Synthetic Data","created":"2026-10-11","cited_by_count":0,"area":null}],"next_cursor":null}