# Perplexity Predicts Protection: Choosing Pretrained Backbones for Worst-Client Fairness in Federated Parameter-Efficient Fine-Tuning

arXiv: https://arxiv.org/abs/2609.23463 · HTML: https://mlsanity.com/paper/2609.23463/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 26 of its 37 references that are in the map.
- First posted on arXiv 20 Sep 2026.

## Builds on (most cited first, 26 of 26)

- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [Communication-Efficient Learning of Deep Networks from Decentralized Data](https://mlsanity.com/paper/1602.05629/) (2016)
- [Character-level Convolutional Networks for Text Classification](https://mlsanity.com/paper/1509.01626/) (2015)
- [Convolutional Neural Networks for Sentence Classification](https://mlsanity.com/paper/1408.5882/) (2014)
- [Domain-Specific Language Model Pretraining for Biomedical Natural Language Processing](https://mlsanity.com/paper/2007.15779/) (2020)
- [Measuring the Effects of Non-Identical Data Distribution for Federated Visual Classification](https://mlsanity.com/paper/1909.06335/) (2019)
- [Ditto: Fair and Robust Federated Learning Through Personalization](https://mlsanity.com/paper/2012.04221/) (2020)
- [Personalized Federated Learning with Moreau Envelopes](https://mlsanity.com/paper/2006.08848/) (2020)
- [Agnostic Federated Learning](https://mlsanity.com/paper/1902.00146/) (2019)
- [Fair Resource Allocation in Federated Learning](https://mlsanity.com/paper/1905.10497/) (2019)
- [BERTweet: A pre-trained language model for English Tweets](https://mlsanity.com/paper/2005.10200/) (2020)
- [How Good is Your Tokenizer? On the Monolingual Performance of Multilingual Language Models](https://mlsanity.com/paper/2012.15613/) (2020)
- [Towards Building the Federated GPT: Federated Instruction Tuning](https://mlsanity.com/paper/2305.05644/) (2023)
- [Improving LoRA in Privacy-preserving Federated Learning](https://mlsanity.com/paper/2403.12313/) (2024)
- [FLoRA: Federated Fine-Tuning Large Language Models with Heterogeneous Low-Rank Adaptations](https://mlsanity.com/paper/2409.05976/) (2024)
- [PubMed 200k RCT: a Dataset for Sequential Sentence Classification in Medical Abstracts](https://mlsanity.com/paper/1710.06071/) (2017)
- [Heterogeneous LoRA for Federated Fine-tuning of On-Device Foundation Models](https://mlsanity.com/paper/2401.06432/) (2024)
- [Stochastic Controlled Averaging for Federated Learning with Communication Compression](https://mlsanity.com/paper/2308.08165/) (2023)
- [Federated Fine-tuning of Large Language Models under Heterogeneous Tasks and Client Resources](https://mlsanity.com/paper/2402.11505/) (2024)
- [Selective Aggregation for Low-Rank Adaptation in Federated Learning](https://mlsanity.com/paper/2410.01463/) (2024)
- [Where to Begin? On the Impact of Pre-Training and Initialization in Federated Learning](https://mlsanity.com/paper/2210.08090/) (2022)
- [pFedLoRA: Model-Heterogeneous Personalized Federated Learning with LoRA Tuning](https://mlsanity.com/paper/2310.13283/) (2023)
- [Non-IID data in Federated Learning: A Survey with Taxonomy, Metrics, Methods, Frameworks and Future Directions](https://mlsanity.com/paper/2411.12377/) (2024)
- [Federated Low-Rank Adaptation for Foundation Models: A Survey](https://mlsanity.com/paper/2505.13502/) (2025)
- [A Survey on Group Fairness in Federated Learning: Challenges, Taxonomy of Solutions and Directions for Future Research](https://mlsanity.com/paper/2410.03855/) (2024)
- [When More Parameters Hurt: Foundation Model Priors Amplify Worst-Client Disparity Under Extreme Federated Heterogeneity](https://mlsanity.com/paper/2605.08992/) (2026)

## Abstract (arXiv)

Federated learning lets multiple parties train a shared model without pooling their data, but a client with far less data than the others can end up poorly served even when the group's average accuracy looks fine. We ask whether the choice of pretrained backbone affects this under LoRA fine-tuning, and whether per-word perplexity on the target text predicts which backbone helps the worst-off client before federated training starts. We ran 313 experiments across three text-classification datasets and three similarly sized backbones (RoBERTa, BERTweet, PubMedBERT), each compared against a task-specific baseline on identical data splits. Lower-perplexity backbones consistently produced larger gains for the worst-performing client, with a rank correlation of -0.87 across nine dataset-backbone pairs; a backbone held out of the analysis confirmed the pattern. Personalization with Ditto recovered only 4-12% of the gap between training alone and full federation, and removing aggregation entirely erased the benefit. A client's update also showed no sign of conflicting with the group's update; the two are close to orthogonal, ruling out one proposed explanation for this failure. Practically: measure perplexity on a sample of task text before choosing a backbone, and do not rely on personalization to protect a data-poor client. We release our code, predictions, and full results for others to test.