# On the Effect of Dropping Layers of Pre-trained Transformer Models

arXiv: https://arxiv.org/abs/2004.03844 · HTML: https://mlsanity.com/paper/2004.03844/

## Key facts

- Cited by 118 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 8 Apr 2020.

## Built on by (most cited first, 30 of 118)

- [The Unreasonable Ineffectiveness of the Deeper Layers](https://mlsanity.com/paper/2403.17887/) (2024)
- [A Fast Post-Training Pruning Framework for Transformers](https://mlsanity.com/paper/2204.09656/) (2022)
- [The Remarkable Robustness of LLMs: Stages of Inference?](https://mlsanity.com/paper/2406.19384/) (2024)
- [Learned Token Pruning for Transformers](https://mlsanity.com/paper/2107.00910/) (2021)
- [Efficient Methods for Natural Language Processing: A Survey](https://mlsanity.com/paper/2209.00099/) (2022)
- [Offsite-Tuning: Transfer Learning without Full Model](https://mlsanity.com/paper/2302.04870/) (2023)
- [FedBiOT: LLM Local Fine-tuning in Federated Learning without Full Model](https://mlsanity.com/paper/2406.17706/) (2024)
- [A Practical Survey on Faster and Lighter Transformers](https://mlsanity.com/paper/2103.14636/) (2021)
- [Harder Tasks Need More Experts: Dynamic Routing in MoE Models](https://mlsanity.com/paper/2403.07652/) (2024)
- [Model Compression and Efficient Inference for Large Language Models: A Survey](https://mlsanity.com/paper/2402.09748/) (2024)
- [Efficient Expert Pruning for Sparse Mixture-of-Experts Language Models: Enhancing Performance and Reducing Inference Costs](https://mlsanity.com/paper/2407.00945/) (2024)
- [Differentiable Subset Pruning of Transformer Heads](https://mlsanity.com/paper/2108.04657/) (2021)
- [Discrete Prompt Compression with Reinforcement Learning](https://mlsanity.com/paper/2308.08758/) (2023)
- [No One Left Behind: Inclusive Federated Learning over Heterogeneous Devices](https://mlsanity.com/paper/2202.08036/) (2022)
- [LGViT: Dynamic Early Exiting for Accelerating Vision Transformer](https://mlsanity.com/paper/2308.00255/) (2023)
- [A Survey of Small Language Models](https://mlsanity.com/paper/2410.20011/) (2024)
- [Gradient-Free Structured Pruning with Unlabeled Data](https://mlsanity.com/paper/2303.04185/) (2023)
- [The Quest for the Right Mediator: Surveying Mechanistic Interpretability Through the Lens of Causal Mediation Analysis](https://mlsanity.com/paper/2408.01416/) (2024)
- [Adapting a Language Model While Preserving its General Knowledge](https://mlsanity.com/paper/2301.08986/) (2023)
- [Fairness-Aware Structured Pruning in Transformers](https://mlsanity.com/paper/2312.15398/) (2023)
- [Discovering Salient Neurons in Deep NLP Models](https://mlsanity.com/paper/2206.13288/) (2022)
- [On the Transformation of Latent Space in Fine-Tuned NLP Models](https://mlsanity.com/paper/2210.12696/) (2022)
- [ECoFLaP: Efficient Coarse-to-Fine Layer-Wise Pruning for Vision-Language Models](https://mlsanity.com/paper/2310.02998/) (2023)
- [OSSCAR: One-Shot Structured Pruning in Vision and Language Models with Combinatorial Optimization](https://mlsanity.com/paper/2403.12983/) (2024)
- [SLMRec: Distilling Large Language Models into Small for Sequential Recommendation](https://mlsanity.com/paper/2405.17890/) (2024)
- [ModuleFormer: Modularity Emerges from Mixture-of-Experts](https://mlsanity.com/paper/2306.04640/) (2023)
- [FedPFT: Federated Proxy Fine-Tuning of Foundation Models](https://mlsanity.com/paper/2404.11536/) (2024)
- [Causal Head Gating: A Framework for Interpreting Roles of Attention Heads in Transformers](https://mlsanity.com/paper/2505.13737/) (2025)
- [Fine-tuning Strategies for Faster Inference using Speech Self-Supervised Models: A Comparative Study](https://mlsanity.com/paper/2303.06740/) (2023)
- [A Comprehensive Survey of Compression Algorithms for Language Models](https://mlsanity.com/paper/2401.15347/) (2024)

## Abstract (arXiv)

Transformer-based NLP models are trained using hundreds of millions or even billions of parameters, limiting their applicability in computationally constrained environments. While the number of parameters generally correlates with performance, it is not clear whether the entire network is required for a downstream task. Motivated by the recent work on pruning and distilling pre-trained models, we explore strategies to drop layers in pre-trained models, and observe the effect of pruning on downstream GLUE tasks. We were able to prune BERT, RoBERTa and XLNet models up to 40%, while maintaining up to 98% of their original performance. Additionally we show that our pruned models are on par with those built using knowledge distillation, both in terms of size and performance. Our experiments yield interesting observations such as, (i) the lower layers are most critical to maintain downstream task performance, (ii) some tasks such as paraphrase detection and sentence similarity are more robust to the dropping of layers, and (iii) models trained using a different objective function exhibit different learning patterns and w.r.t the layer dropping.