Beyond Initialization Loss: A Systematic Study of Token Embedding Initialization Strategies for LLM Vocabulary Extension
Raviraj Joshi, Utkarsh Vaidya, Sanjay Singh Chauhan, Niranjan Wartikar
cs.CL, cs.LG
Submitted: 2026-08-04
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Llama-3-Nanda-10B-Chat: An Open Generative Large Language Model for Hindi
- Efficient and Effective Text Encoding for Chinese LLaMA and Alpaca
- Continual Pre-Training for Cross-Lingual LLM Adaptation: Enhancing Japanese Language Capabilities
- Adapting Multilingual LLMs to Low-Resource Languages using Continued Pre-training and Synthetic Corpus
- MuRIL: Multilingual Representations for Indian Languages
- AdaptiVocab: Enhancing LLM Efficiency in Focused Domains through Lightweight Vocabulary Adaptation
- Neural Machine Translation of Rare Words with Subword Units
- Gemma 2: Improving Open Language Models at a Practical Size
- RedWhale: An Adapted Korean LLM Through Efficient Continual Pretraining
Related papers
- Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
- Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
- Subliminal Steering: Stronger Encoding of Hidden Signals
- MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
- The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
- Untangling the Mechanisms of Misleading Context in Medical Question Answering