# Memory-Efficient Federated Fine-Tuning of Large Language Models via Layer Pruning

arXiv: https://arxiv.org/abs/2508.17209 · HTML: https://mlsanity.com/paper/2508.17209/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 24 Aug 2025.

## Built on by (most cited first, 5 of 5)

- [TSEmbed: Unlocking Task Scaling in Universal Multimodal Embeddings](https://mlsanity.com/paper/2603.04772/) (2026)
- [When Agents "Misremember" Collectively: Exploring the Mandela Effect in LLM-based Multi-Agent Systems](https://mlsanity.com/paper/2602.00428/) (2026)
- ["I See What You Did There": Can Large Vision-Language Models Understand Multimodal Puns?](https://mlsanity.com/paper/2604.05930/) (2026)
- [Spotlight and Shadow: Attention-Guided Dual-Anchor Introspective Decoding for MLLM Hallucination Mitigation](https://mlsanity.com/paper/2604.10071/) (2026)
- [CoEM: Empowering Long-Context Reasoning with Commit-on-Evidence Memory](https://mlsanity.com/paper/2609.36935/) (2026)

## Abstract (arXiv)

Federated fine-tuning enables privacy-preserving Large Language Model (LLM) adaptation, but its high memory cost limits participation from resource-constrained devices. We propose FedPruner, an innovative federated fine-tuning paradigm that tackles this via intelligent layer pruning. FedPruner flexibly prunes the global model, creating personalized submodels based on device memory constraints. It employs a macro-micro synergistic pruning framework: a macro-level functionality-driven layer orchestration mechanism groups layers, while a micro-level importance-aware layer selection strategy prunes within groups to build device-specific submodels. We further introduce a fine-grained variant that independently prunes Multi-Head Attention and Feed-Forward Network components to precisely preserve critical architectural elements. Extensive experimental results demonstrate that FedPruner significantly outperforms state-of-the-art approaches, achieving up to a 1.98% improvement in average model accuracy while reducing peak memory usage by 75%.