# Classifier Language Models: Unifying Sparse Finetuning and Adaptive Tokenization for Specialized Classification Tasks

arXiv: https://arxiv.org/abs/2508.08635 · HTML: https://mlsanity.com/paper/2508.08635/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 18 of its 32 references that are in the map.
- First posted on arXiv 12 Aug 2025.

## Builds on (most cited first, 18 of 18)

- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [Parameter-Efficient Transfer Learning for NLP](https://mlsanity.com/paper/1902.00751/) (2019)
- [Prefix-Tuning: Optimizing Continuous Prompts for Generation](https://mlsanity.com/paper/2101.00190/) (2021)
- [The Power of Scale for Parameter-Efficient Prompt Tuning](https://mlsanity.com/paper/2104.08691/) (2021)
- [Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone](https://mlsanity.com/paper/2404.14219/) (2024)
- [BitFit: Simple Parameter-efficient Fine-tuning for Transformer-based Masked Language-models](https://mlsanity.com/paper/2106.10199/) (2021)
- [AdapterFusion: Non-Destructive Task Composition for Transfer Learning](https://mlsanity.com/paper/2005.00247/) (2020)
- [How to Fine-Tune BERT for Text Classification?](https://mlsanity.com/paper/1905.05583/) (2019)
- [On the Stability of Fine-tuning BERT: Misconceptions, Explanations, and Strong Baselines](https://mlsanity.com/paper/2006.04884/) (2020)
- [DyLoRA: Parameter Efficient Tuning of Pre-trained Models using Dynamic Search-Free Low-Rank Adaptation](https://mlsanity.com/paper/2210.07558/) (2022)
- [LISA: Layerwise Importance Sampling for Memory-Efficient Large Language Model Fine-Tuning](https://mlsanity.com/paper/2403.17919/) (2024)
- [Small and Practical BERT Models for Sequence Labeling](https://mlsanity.com/paper/1909.00100/) (2019)
- [Improved Fine-Tuning by Better Leveraging Pre-Training Data](https://mlsanity.com/paper/2111.12292/) (2021)
- [Towards Understanding Multi-Task Learning (Generalization) of LLMs via Detecting and Exploring Task-Specific Neurons](https://mlsanity.com/paper/2407.06488/) (2024)
- [AMBERT: A Pre-trained Language Model with Multi-Grained Tokenization](https://mlsanity.com/paper/2008.11869/) (2020)
- [Bayesian Optimization of Text Representations](https://mlsanity.com/paper/1503.00693/) (2015)
- [Unigram-Normalized Perplexity as a Language Model Performance Measure with Different Vocabulary Sizes](https://mlsanity.com/paper/2011.13220/) (2020)

## Abstract (arXiv)

Semantic text classification requires the understanding of the contextual significance of specific tokens rather than surface-level patterns or keywords (as in rule-based or statistical text classification), making large language models (LLMs) well-suited for this task. However, semantic classification applications in industry, like customer intent detection or semantic role labeling, tend to be highly specialized. They require annotation by domain experts in contrast to general-purpose corpora for pretraining. Further, they typically require high inference throughputs which limits the model size from latency and cost perspectives. Thus, for a range of specialized classification tasks, the preferred solution is to develop customized classifiers by finetuning smaller language models (e.g., mini-encoders, small language models). In this work, we develop a token-driven sparse finetuning strategy to adapt small language models to specialized classification tasks. We identify and finetune a small sensitive subset of model parameters by leveraging task-specific token constructs in the finetuning dataset, while leaving most of the pretrained weights unchanged. Unlike adapter approaches such as low rank adaptation (LoRA), we do not introduce additional parameters to the model. Our approach identifies highly relevant semantic tokens (case study in the Appendix) and outperforms end-to-end finetuning, LoRA, layer selection, and prefix tuning on five diverse semantic classification tasks. We achieve greater stability and half the training costs vs. end-to-end finetuning.