A Note on Shumailov et al. (2024): `AI Models Collapse When Trained on Recursively Generated Data'

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autore principale:	Borji, Ali
Natura:	Preprint
Pubblicazione:	2024
Soggetti:	Machine Learning Artificial Intelligence
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

ForTIFAI: Fending Off Recursive Training Induced Failure for AI Model Collapse
di: Shabgahi, Soheil Zibakhsh, et al.
Pubblicazione: (2025)

When Privacy Isn't Synthetic: Hidden Data Leakage in Generative AI Models
di: Mustaqim, S. M., et al.
Pubblicazione: (2025)

Is Model Collapse Inevitable? Breaking the Curse of Recursion by Accumulating Real and Synthetic Data
di: Gerstgrasser, Matthias, et al.
Pubblicazione: (2024)

Multi-modal Synthetic Data Training and Model Collapse: Insights from VLMs and Diffusion Models
di: Hu, Zizhao, et al.
Pubblicazione: (2025)

Curated Synthetic Data Doesn't Have to Collapse: A Theoretical Study of Generative Retraining with Pluralistic Preferences
di: Falahati, Ali, et al.
Pubblicazione: (2026)

Cyborg Data: Merging Human with AI Generated Training Data
di: North, Kai, et al.
Pubblicazione: (2025)

When Data Is Scarce: Scaling Sparse Language Models with Repeated Training
di: Wu, Boqian, et al.
Pubblicazione: (2026)

Escaping Collapse: The Strength of Weak Data for Large Language Model Training
di: Amin, Kareem, et al.
Pubblicazione: (2025)

How Bad is Training on Synthetic Data? A Statistical Analysis of Language Model Collapse
di: Seddik, Mohamed El Amine, et al.
Pubblicazione: (2024)

The Geometry of Alignment Collapse: When Fine-Tuning Breaks Safety
di: Springer, Max, et al.
Pubblicazione: (2026)

A Theoretical Perspective: How to Prevent Model Collapse in Self-consuming Training Loops
di: Fu, Shi, et al.
Pubblicazione: (2025)

A Note on Statistically Accurate Tabular Data Generation Using Large Language Models
di: Sidorenko, Andrey
Pubblicazione: (2025)

Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training
di: Wang, Kevin, et al.
Pubblicazione: (2026)

Dominating vs. Dominated: Generative Collapse in Diffusion Models
di: Jeong, Hayeon, et al.
Pubblicazione: (2025)

The Curse of Recursion: Training on Generated Data Makes Models Forget
di: Shumailov, Ilia, et al.
Pubblicazione: (2023)

Enhancing Pre-Trained Model-Based Class-Incremental Learning through Neural Collapse
di: He, Kun, et al.
Pubblicazione: (2025)

Qualitative Failures of Image Generation Models and Their Application in Detecting Deepfakes
di: Borji, Ali
Pubblicazione: (2023)

Collapse or Thrive? Perils and Promises of Synthetic Data in a Self-Generating World
di: Kazdan, Joshua, et al.
Pubblicazione: (2024)

KODA: A Data-Driven Recursive Model for Time Series Forecasting and Data Assimilation using Koopman Operators
di: Singh, Ashutosh, et al.
Pubblicazione: (2024)

Beyond Model Collapse: Scaling Up with Synthesized Data Requires Verification
di: Feng, Yunzhen, et al.
Pubblicazione: (2024)

When Safety Geometry Collapses: Fine-Tuning Vulnerabilities in Agentic Guard Models
di: Hossain, Ismail, et al.
Pubblicazione: (2026)

The Alignment Game: A Theory of Long-Horizon Alignment Through Recursive Curation
di: Falahati, Ali, et al.
Pubblicazione: (2025)

Zipping the Thought: When and How Compressed Reasoning Data Works in LLM Post-Training
di: Matsutani, Kohsei, et al.
Pubblicazione: (2026)

When Attention Collapses: How Degenerate Layers in LLMs Enable Smaller, Stronger Models
di: Sanyal, Sunny, et al.
Pubblicazione: (2024)

Recursive Decomposition with Dependencies for Generic Divide-and-Conquer Reasoning
di: Hernández-Gutiérrez, Sergio, et al.
Pubblicazione: (2025)

The Achilles Heel of AI: Fundamentals of Risk-Aware Training Data for High-Consequence Models
di: Cook, Dave, et al.
Pubblicazione: (2025)

When Attention Collapses: Residual Evidence Modeling for Compositional Inference
di: Houba, Niklas
Pubblicazione: (2026)

Test-time Adaptation of Tiny Recursive Models
di: McGovern, Ronan Killian
Pubblicazione: (2025)

Accelerating Training Speed of Tiny Recursive Models with Curriculum Guided Adaptive Recursion
di: Qasim, Kaleem Ullah, et al.
Pubblicazione: (2025)

Generative Models for Synthetic Data: Transforming Data Mining in the GenAI Era
di: Li, Dawei, et al.
Pubblicazione: (2025)

Model Collapse Demystified: The Case of Regression
di: Dohmatob, Elvis, et al.
Pubblicazione: (2024)

How to Synthesize Text Data without Model Collapse?
di: Zhu, Xuekai, et al.
Pubblicazione: (2024)

When AI Eats Itself: On the Caveats of AI Autophagy
di: Xing, Xiaodan, et al.
Pubblicazione: (2024)

Downstream Task-Oriented Generative Model Selections on Synthetic Data Training for Fraud Detection Models
di: Cheng, Yinan, et al.
Pubblicazione: (2024)

When Reasoning Hurts: Source-Aware Evaluation of Frontier LLMs for Clinical SOAP Note Generation
di: Faisal, Faizan
Pubblicazione: (2026)

Grokking and Generalization Collapse: Insights from \texttt{HTSR} theory
di: Prakash, Hari K., et al.
Pubblicazione: (2025)

Distribution Fitting for Combating Mode Collapse in Generative Adversarial Networks
di: Gong, Yanxiang, et al.
Pubblicazione: (2022)

Scaling with Collapse: Efficient and Predictable Training of LLM Families
di: Bergsma, Shane, et al.
Pubblicazione: (2025)

Critical Windows of Complexity Control: When Transformers Decide to Reason or Memorize
di: Ali, Sarwan
Pubblicazione: (2026)

SnapE -- Training Snapshot Ensembles of Link Prediction Models
di: Shaban, Ali, et al.
Pubblicazione: (2024)