Curated Synthetic Data Doesn't Have to Collapse: A Theoretical Study of Generative Retraining with Pluralistic Preferences

Documenti analoghi

• The Alignment Game: A Theory of Long-Horizon Alignment Through Recursive Curation
  di: Falahati, Ali, et al.
  Pubblicazione: (2025)
• Disentangled Structural and Featural Representation for Task-Agnostic Graph Valuation
  di: Falahati, Ali, et al.
  Pubblicazione: (2024)
• A Median Perspective on Unlabeled Data for Out-of-Distribution Detection
  di: Abbas, Momin, et al.
  Pubblicazione: (2025)
• SentenceKV: Efficient LLM Inference via Sentence-Level Semantic KV Caching
  di: Zhu, Yuxuan, et al.
  Pubblicazione: (2025)
• Machine Unlearning Doesn't Do What You Think: Lessons for Generative AI Policy and Research
  di: Cooper, A. Feder, et al.
  Pubblicazione: (2024)