When and How Human Curation Backfires: Preference Alignment under Multi-Model Self-Consuming Loop

• URL: http://arxiv.org/abs/2605.29267v1
• Date: Thu, 28 May 2026 02:36:57 GMT
• Title: When and How Human Curation Backfires: Preference Alignment under Multi-Model Self-Consuming Loop
• Abstract summary: Self-consuming training can lead to model collapse, divergence, or bias amplification.<n>Recent work shows that incorporating human curation into the loop can steer a self-consuming model toward human-aligned behavior.<n>This paper studies self-consuming training in the multi-model regime.
• Score: 13.054822893327058
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Foundation models are increasingly trained on synthetic data generated by prior model iterations rather than exclusively on real data. This self-consuming training paradigm can lead to model collapse, divergence, or bias amplification. Recent work (Ferbach et al., 2024) shows that incorporating human curation into the loop can steer a self-consuming model toward human-aligned behavior, but these analyses focus on a single, isolated model that solely consumes its own outputs. In practice, however, models often interact and train on input-output pairs produced by other models. This paper studies self-consuming training in the multi-model regime. We first formalize a framework for interacting self-consuming models and characterize when the resulting dynamical system converges to a stable point. We then examine how human curation of one model affects its own alignment (self-influence) and how such effects propagate to other models (cross-influence). Unlike isolated settings where human curation always enhances model alignment, we show that cross-model interactions can dampen or even invert this effect, ultimately degrading long-term alignment.

Related papers

• When Does Model Collapse Occur in Structured Interactive Learning? [9.385946237424612]
  We investigate the performance of generative models in an interactive learning environment with general interaction patterns.<n>We show that the occurrence of model collapse depends critically on the topology of the interaction graph.
  arXiv  Detail & Related papers  (2026-05-19T17:41:09Z)
• Multiagent Finetuning: Self Improvement with Diverse Reasoning Chains [114.76612918465948]
  Large language models (LLMs) have achieved remarkable performance in recent years but are fundamentally limited by the underlying training data.<n>We propose a complementary approach towards self-improvement where finetuning is applied to a multiagent society of language models.
  arXiv  Detail & Related papers  (2025-01-10T04:35:46Z)
• Ablation Based Counterfactuals [7.481286710933861]
  Ablation Based Counterfactuals (ABC) is a method of performing counterfactual analysis that relies on model ablation rather than model retraining. We demonstrate how we can construct a model like this using an ensemble of diffusion models. We then use this model to study the limits of training data attribution by enumerating full counterfactual landscapes.
  arXiv  Detail & Related papers  (2024-06-12T06:22:51Z)
• Unified Dynamic Scanpath Predictors Outperform Individually Trained Neural Models [18.327960366321655]
  We develop a deep learning-based social cue integration model for saliency prediction to predict scanpaths in videos.<n>We evaluate our approach on gaze of dynamic social scenes, observed under the free-viewing condition.<n>Results indicate that a single unified model, trained on all the observers' scanpaths, performs on par or better than individually trained models.
  arXiv  Detail & Related papers  (2024-05-05T13:15:11Z)
• How Bad is Training on Synthetic Data? A Statistical Analysis of Language Model Collapse [9.59833542807268]
  Model collapse occurs when new models are trained on synthetic data generated from previously trained models. We show that model collapse cannot be avoided when training solely on synthetic data. We estimate a maximal amount of synthetic data below which model collapse can eventually be avoided.
  arXiv  Detail & Related papers  (2024-04-07T22:15:13Z)
• Is Model Collapse Inevitable? Breaking the Curse of Recursion by Accumulating Real and Synthetic Data [49.73114504515852]
  We show that replacing the original real data by each generation's synthetic data does indeed tend towards model collapse. We demonstrate that accumulating the successive generations of synthetic data alongside the original real data avoids model collapse.
  arXiv  Detail & Related papers  (2024-04-01T18:31:24Z)
• Towards Theoretical Understandings of Self-Consuming Generative Models [56.84592466204185]
  This paper tackles the emerging challenge of training generative models within a self-consuming loop. We construct a theoretical framework to rigorously evaluate how this training procedure impacts the data distributions learned by future models. We present results for kernel density estimation, delivering nuanced insights such as the impact of mixed data training on error propagation.
  arXiv  Detail & Related papers  (2024-02-19T02:08:09Z)
• On the Stability of Iterative Retraining of Generative Models on their own Data [56.153542044045224]
  We study the impact of training generative models on mixed datasets. We first prove the stability of iterative training under the condition that the initial generative models approximate the data distribution well enough. We empirically validate our theory on both synthetic and natural images by iteratively training normalizing flows and state-of-the-art diffusion models.
  arXiv  Detail & Related papers  (2023-09-30T16:41:04Z)
• Dataless Knowledge Fusion by Merging Weights of Language Models [47.432215933099016]
  Fine-tuning pre-trained language models has become the prevalent paradigm for building downstream NLP models.<n>This creates a barrier to fusing knowledge across individual models to yield a better single model.<n>We propose a dataless knowledge fusion method that merges models in their parameter space.
  arXiv  Detail & Related papers  (2022-12-19T20:46:43Z)
• Are Neural Topic Models Broken? [81.15470302729638]
  We study the relationship between automated and human evaluation of topic models. We find that neural topic models fare worse in both respects compared to an established classical method.
  arXiv  Detail & Related papers  (2022-10-28T14:38:50Z)
• Is Automated Topic Model Evaluation Broken?: The Incoherence of Coherence [62.826466543958624]
  We look at the standardization gap and the validation gap in topic model evaluation. Recent models relying on neural components surpass classical topic models according to these metrics. We use automatic coherence along with the two most widely accepted human judgment tasks, namely, topic rating and word intrusion.
  arXiv  Detail & Related papers  (2021-07-05T17:58:52Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.