Self-Consuming Generative Models with Co-Evolving Human Preferences
cs.LG
Submitted: 2026-10-07
Updated: 2026-10-07
Code: https://github.com/osu-srml/Model-Reward-Coevolution
Terminology
Sources
- Self-Consuming Generative Models with Curated Data Provably Optimize Human Preferences
- Explaining and Preventing Alignment Collapse in Iterative RLHF
- Distributionally Robust Performative Optimization
- Making Decisions under Outcome Performativity
- Clipped SGD Algorithms for Performative Prediction: Tight Bounds for Clipping Bias and Remedies
- Two-timescale Derivative Free Optimization for Performative Prediction with Markovian Data
- Improved Denoising Diffusion Probabilistic Models
- Rate of Model Collapse in Recursive Training
- Observations and Remedies for Large Language Model Bias in Self-Consuming Performative Loop
- Self-Consuming Generative Models with Adversarially Curated Data
- SPRINT: Stochastic Performative Prediction With Variance Reduction
- Character-level Convolutional Networks for Text Classification
- When and How Human Curation Backfires: Preference Alignment under Multi-Model Self-Consuming Loop
- Convergence and Stability Analysis of Self-Consuming Generative Models with Heterogeneous Human Curation
- Optimizing the Performative Risk under Weak Convexity Assumptions
- ProFL: Performative Robust Optimal Federated Learning
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks