# Optimizing Korean-Centric LLMs via Token Pruning

arXiv: https://arxiv.org/abs/2604.16235 · HTML: https://mlsanity.com/paper/2604.16235/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in KV cache optimization, one of 12 directions in Efficient LLM inference and serving.
- Builds on 11 of its 17 references that are in the map.
- First posted on arXiv 17 Apr 2026.

## Placed in

- Efficient LLM inference and serving / [KV cache optimization](https://mlsanity.com/inference/kv-cache-optimization/)

## Builds on (most cited first, 11 of 11)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [xCOMET: Transparent Machine Translation Evaluation through Fine-grained Error Detection](https://mlsanity.com/paper/2310.10482/) (2023)
- [KMMLU: Measuring Massive Multitask Language Understanding in Korean](https://mlsanity.com/paper/2402.11548/) (2024)
- [WMT24++: Expanding the Language Coverage of WMT24 to 55 Languages & Dialects](https://mlsanity.com/paper/2502.12404/) (2025)
- [Token Pruning in Multimodal Large Language Models: Are We Solving the Right Problem?](https://mlsanity.com/paper/2502.11501/) (2025)
- [Understanding and Mitigating Language Confusion in LLMs](https://mlsanity.com/paper/2406.20052/) (2024)
- [HAE-RAE Bench: Evaluation of Korean Knowledge in Language Models](https://mlsanity.com/paper/2309.02706/) (2023)
- [Seed-X: Building Strong Multilingual Translation LLM with 7B Parameters](https://mlsanity.com/paper/2507.13618/) (2025)
- [RedWhale: An Adapted Korean LLM Through Efficient Continual Pretraining](https://mlsanity.com/paper/2408.11294/) (2024)
- [Dynamic Vocabulary Pruning in Early-Exit LLMs](https://mlsanity.com/paper/2410.18952/) (2024)

## Built on by (most cited first, 1 of 1)

- [Language Models are not Equally Robust to Non-Canonical Tokenization across Languages](https://mlsanity.com/paper/2607.26831/) (2026)

## Abstract (arXiv)

This paper presents a systematic benchmark of state-of-the-art multilingual large language models (LLMs) adapted via token pruning - a compression technique that eliminates tokens and embedding parameters corresponding to languages irrelevant to the target application. Focusing on Korean-centric natural language processing (NLP) tasks, we evaluate architectures including Qwen3, Gemma-3, Llama-3, and Aya across three vocabulary configurations: Original, English-Korean (EnKo), and English-Korean-Chinese (EnKoZh). Performance is assessed using established benchmarks for general aptitude, cultural literacy, instruction following, and machine translation. Our findings indicate that token pruning significantly improves generation stability by eliminating language confusion, and in the case of machine translation, frequently enhances performance on Korean-specific tasks. While instruction-following capabilities display architecture-dependent variance linked to latent cross-lingual representations, the significant reduction in vocabulary size validates token pruning as a highly effective optimization strategy for memory-constrained, domain-specific deployments, despite modest gains in inference latency.