# Self-Consuming Generative Models with Co-Evolving Human Preferences

arXiv: https://arxiv.org/abs/2610.09415 · HTML: https://mlsanity.com/paper/2610.09415/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 17 of its 62 references that are in the map.
- First posted on arXiv 7 Oct 2026.

## Builds on (most cited first, 17 of 17)

- [Improved Denoising Diffusion Probabilistic Models](https://mlsanity.com/paper/2102.09672/) (2021)
- [Character-level Convolutional Networks for Text Classification](https://mlsanity.com/paper/1509.01626/) (2015)
- [Self-Consuming Generative Models with Curated Data Provably Optimize Human Preferences](https://mlsanity.com/paper/2407.09499/) (2024)
- [Making Decisions under Outcome Performativity](https://mlsanity.com/paper/2210.01745/) (2022)
- [Rate of Model Collapse in Recursive Training](https://mlsanity.com/paper/2412.17646/) (2024)
- [How LLMs Distort Our Written Language](https://mlsanity.com/paper/2603.18161/) (2026)
- [Two-timescale Derivative Free Optimization for Performative Prediction with Markovian Data](https://mlsanity.com/paper/2310.05792/) (2023)
- [Self-Consuming Generative Models with Adversarially Curated Data](https://mlsanity.com/paper/2505.09768/) (2025)
- [Optimizing the Performative Risk under Weak Convexity Assumptions](https://mlsanity.com/paper/2209.00771/) (2022)
- [Clipped SGD Algorithms for Performative Prediction: Tight Bounds for Clipping Bias and Remedies](https://mlsanity.com/paper/2404.10995/) (2024)
- [Explaining and Preventing Alignment Collapse in Iterative RLHF](https://mlsanity.com/paper/2605.04266/) (2026)
- [Convergence and Stability Analysis of Self-Consuming Generative Models with Heterogeneous Human Curation](https://mlsanity.com/paper/2511.09002/) (2025)
- [Distributionally Robust Performative Optimization](https://mlsanity.com/paper/2407.01344/) (2024)
- [ProFL: Performative Robust Optimal Federated Learning](https://mlsanity.com/paper/2410.18075/) (2024)
- [SPRINT: Stochastic Performative Prediction With Variance Reduction](https://mlsanity.com/paper/2509.17304/) (2025)
- [Observations and Remedies for Large Language Model Bias in Self-Consuming Performative Loop](https://mlsanity.com/paper/2601.05184/) (2026)
- [When and How Human Curation Backfires: Preference Alignment under Multi-Model Self-Consuming Loop](https://mlsanity.com/paper/2605.29267/) (2026)

## Abstract (arXiv)

Generative models are increasingly trained in self-consuming iterative loops, where users curate preferred samples from model-generated candidates and the curated samples are used to train future generations of the model. Prior work has largely assumed fixed user preferences, but in practice exposure to model outputs gradually reshapes what users perceive as desirable, creating a feedback loop in which model distributions and user preferences co-evolve. We take a first step toward understanding the long-term behavior of such coupled dynamics. We show that when training relies entirely on user-curated synthetic data, iterative curation amplifies initial biases and drives the system toward one of multiple singleton equilibria in which the instance holding an initial advantage eventually dominates. In contrast, injecting reference data into training at a sufficiently large rate fundamentally changes the dynamics and yields a unique globally attracting equilibrium. Building on this insight, we study how reference-data injection can be used to control long-term outcomes, and propose an efficient algorithm that jointly selects a reference distribution and its mixing weight to steer the coupled system toward equilibria that preserve desired attributes while minimizing data collection costs.