# DELIA: Diversity-Enhanced Learning for Instruction Adaptation in Large Language Models

arXiv: https://arxiv.org/abs/2408.10841 · HTML: https://mlsanity.com/paper/2408.10841/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 10 of its 14 references that are in the map.
- First posted on arXiv 19 Aug 2024.

## Builds on (most cited first, 10 of 10)

- [LIMA: Less Is More for Alignment](https://mlsanity.com/paper/2305.11206/) (2023)
- [Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formatting](https://mlsanity.com/paper/2310.11324/) (2023)
- [NEFTune: Noisy Embeddings Improve Instruction Finetuning](https://mlsanity.com/paper/2310.05914/) (2023)
- [The unreasonable effectiveness of few-shot learning for machine translation](https://mlsanity.com/paper/2302.01398/) (2023)
- [Dictionary-based Phrase-level Prompting of Large Language Models for Machine Translation](https://mlsanity.com/paper/2302.07856/) (2023)
- [Injecting New Knowledge into Large Language Models via Supervised Fine-Tuning](https://mlsanity.com/paper/2404.00213/) (2024)
- [Two-stage LLM Fine-tuning with Less Specialization and More Generalization](https://mlsanity.com/paper/2211.00635/) (2022)
- [SELF-GUIDE: Better Task-Specific Instruction Following via Self-Synthetic Finetuning](https://mlsanity.com/paper/2407.12874/) (2024)
- [Reflection-Tuning: Data Recycling Improves LLM Instruction-Tuning](https://mlsanity.com/paper/2310.11716/) (2023)
- [Domain Adversarial Fine-Tuning as an Effective Regularizer](https://mlsanity.com/paper/2009.13366/) (2020)

## Abstract (arXiv)

Although instruction tuning is widely used to adjust behavior in Large Language Models (LLMs), extensive empirical evidence and research indicates that it is primarily a process where the model fits to specific task formats, rather than acquiring new knowledge or capabilities. We propose that this limitation stems from biased features learned during instruction tuning, which differ from ideal task-specfic features, leading to learn less underlying semantics in downstream tasks. However, ideal features are unknown and incalculable, constraining past work to rely on prior knowledge to assist reasoning or training, which limits LLMs' capabilities to the developers' abilities, rather than data-driven scalable learning. In our paper, through our novel data synthesis method, DELIA (Diversity-Enhanced Learning for Instruction Adaptation), we leverage the buffering effect of extensive diverse data in LLMs training to transform biased features in instruction tuning into approximations of ideal features, without explicit prior ideal features. Experiments show DELIA's better performance compared to common instruction tuning and other baselines. It outperforms common instruction tuning by 17.07%-33.41% on Icelandic-English translation bleurt score (WMT-21 dataset, gemma-7b-it) and improves accuracy by 36.1% on formatted text generation (Llama2-7b-chat). Notably, among knowledge injection methods we've known, DELIA uniquely align the internal representations of new special tokens with their prior semantics.