# CascadeBERT: Accelerating Inference of Pre-trained Language Models via Calibrated Complete Models Cascade

arXiv: https://arxiv.org/abs/2012.14682 · HTML: https://mlsanity.com/paper/2012.14682/

## Key facts

- Cited by 47 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 29 Dec 2020.

## Built on by (most cited first, 30 of 47)

- [Transformer Feed-Forward Layers Build Predictions by Promoting Concepts in the Vocabulary Space](https://mlsanity.com/paper/2203.14680/) (2022)
- [Confident Adaptive Language Modeling](https://mlsanity.com/paper/2207.07061/) (2022)
- [MoEfication: Transformer Feed-forward Layers are Mixtures of Experts](https://mlsanity.com/paper/2110.01786/) (2021)
- [LLM Inference Unveiled: Survey and Roofline Model Insights](https://mlsanity.com/paper/2402.16363/) (2024)
- [Towards Efficient Generative Large Language Model Serving: A Survey from Algorithms to Systems](https://mlsanity.com/paper/2312.15234/) (2023)
- [Not All Layers of LLMs Are Necessary During Inference](https://mlsanity.com/paper/2403.02181/) (2024)
- [SkipDecode: Autoregressive Skip Decoding with Batching and Caching for Efficient LLM Inference](https://mlsanity.com/paper/2307.02628/) (2023)
- [A Unified Approach to Routing and Cascading for LLMs](https://mlsanity.com/paper/2410.10347/) (2024)
- [Towards Efficient NLP: A Standard Evaluation and A Strong Baseline](https://mlsanity.com/paper/2110.07038/) (2021)
- [Model Cascading: Towards Jointly Improving Efficiency and Accuracy of NLP Systems](https://mlsanity.com/paper/2210.05528/) (2022)
- [You Need Multiple Exiting: Dynamic Early Exiting for Accelerating Unified Vision Language Model](https://mlsanity.com/paper/2211.11152/) (2022)
- [A Simple Hash-Based Early Exiting Approach For Language Understanding and Generation](https://mlsanity.com/paper/2203.01670/) (2022)
- [System-1.5 Reasoning: Traversal in Language and Latent Spaces with Dynamic Shortcuts](https://mlsanity.com/paper/2505.18962/) (2025)
- [A Survey on Dynamic Neural Networks for Natural Language Processing](https://mlsanity.com/paper/2202.07101/) (2022)
- [FFN-SkipLLM: A Hidden Gem for Autoregressive Decoding with Adaptive Feed Forward Skipping](https://mlsanity.com/paper/2404.03865/) (2024)
- [Post-Abstention: Towards Reliably Re-Attempting the Abstained Instances in QA](https://mlsanity.com/paper/2305.01812/) (2023)
- [Online Cascade Learning for Efficient Inference over Streams](https://mlsanity.com/paper/2402.04513/) (2024)
- [Efficient Prompting via Dynamic In-Context Learning](https://mlsanity.com/paper/2305.11170/) (2023)
- [Agreement-Based Cascading for Efficient Inference](https://mlsanity.com/paper/2407.02348/) (2024)
- [TangoBERT: Reducing Inference Cost by using Cascaded Architecture](https://mlsanity.com/paper/2204.06271/) (2022)
- [Alternating Updates for Efficient Transformers](https://mlsanity.com/paper/2301.13310/) (2023)
- [ConsistentEE: A Consistent and Hardness-Guided Early Exiting Method for Accelerating Language Models Inference](https://mlsanity.com/paper/2312.11882/) (2023)
- [BabyBear: Cheap inference triage for expensive language models](https://mlsanity.com/paper/2205.11747/) (2022)
- [When Neural Code Completion Models Size up the Situation: Attaining Cheaper and Faster Completion through Dynamic Model Inference](https://mlsanity.com/paper/2401.09964/) (2024)
- [Can LLMs get help from other LLMs without revealing private information?](https://mlsanity.com/paper/2404.01041/) (2024)
- [Communication Efficient Federated Learning for Multilingual Neural Machine Translation with Adapter](https://mlsanity.com/paper/2305.12449/) (2023)
- [Towards an On-device Agent for Text Rewriting](https://mlsanity.com/paper/2308.11807/) (2023)
- [Hierarchical Skip Decoding for Efficient Autoregressive Text Generation](https://mlsanity.com/paper/2403.14919/) (2024)
- [Finding the SWEET Spot: Analysis and Improvement of Adaptive Inference in Low Resource Settings](https://mlsanity.com/paper/2306.02307/) (2023)
- [Bridging the Gap between Decision and Logits in Decision-based Knowledge Distillation for Pre-trained Language Models](https://mlsanity.com/paper/2306.08909/) (2023)

## Abstract (arXiv)

Dynamic early exiting aims to accelerate the inference of pre-trained language models (PLMs) by emitting predictions in internal layers without passing through the entire model. In this paper, we empirically analyze the working mechanism of dynamic early exiting and find that it faces a performance bottleneck under high speed-up ratios. On one hand, the PLMs' representations in shallow layers lack high-level semantic information and thus are not sufficient for accurate predictions. On the other hand, the exiting decisions made by internal classifiers are unreliable, leading to wrongly emitted early predictions. We instead propose a new framework for accelerating the inference of PLMs, CascadeBERT, which dynamically selects proper-sized and complete models in a cascading manner, providing comprehensive representations for predictions. We further devise a difficulty-aware objective, encouraging the model to output the class probability that reflects the real difficulty of each instance for a more reliable cascading mechanism. Experimental results show that CascadeBERT can achieve an overall 15% improvement under 4× speed-up compared with existing dynamic early exiting methods on six classification tasks, yielding more calibrated and accurate predictions.