Mixtures-of-Experts models overfit more to repeated data
Data Scarcity and Model Sparsity: Mixtures-of-Experts Overfit More to Repeated Data
This matters to developers working with large language models because it highlights a critical trade-off: MoE models, while more efficient, are more vulnerable to overfitting when training data is repeated, potentially leading to worse performance in real-world applications.
A new study finds that Mixture-of-Experts (MoE) models, which are popular for their efficiency, suffer more performance degradation when trained on repeated data compared to dense models. For models ranging from 80M to 1B parameters, MoEs begin to suffer at 4x data repetition, while dense models can handle 8x with minimal issues. The study, which tested regularization methods like dropout, found that while some help, none fully match the performance of using all-unique data. It also identified that expert specialization correlates with overfitting to repeated data.
Data Scarcity and Model Sparsity: Mixtures-of-Experts Overfit More to Repeated Data
As the supply of human-written text is exhausted, it has become standard practice to repeat language model training data. Prior work has studied data repetition for densely activated Transformers, but the effects of data repetition remains largely unexplored for recently dominant sparse architectures such as Mixture-of-Experts (MoE), despite their increased compute efficiency. We vary data repetition rates across single- and multi-domain data mixes, and across MoE settings, including expert count and granularity. We consistently find, for models ranging from 80M to 1B active (8.5B total) parameters, that MoEs degrade more rapidly under data repetition. This effect increases with sparsity, dictated by total rather than active parameters. While 80M dense models can repeat data over 8x with minimal degradation, MoEs instead begin to suffer at 4x, and deteriorate rapidly, ceding their performance benefits in all-unique data settings to underperform dense models after 32x. We experiment with existing regularization methods as a potential remedy. We find that some methods, such as dropout, can mitigate overfitting. In particular, with strong masking-based regularization, MoEs are able to outperform dense models even when data is repeated more than 64 times. However, no method fully matches the performance of all-unique training data. Finally, we analyze internal mechanisms correlated with MoE overfitting in high repetition regimes, and find that MoE routing universally stabilizes early in training, and that expert specialization correlates with overfitting to repeated data. In sum, our work addresses the adverse interactions between sparsity and data repetition: we present evidence for the core mechanisms of overfitting and its potential remediation, and suggest promising avenues for future methods to reduce over-specialization in model parameters by disrupting memorization patterns.
