LLM Adaptation

Adapting large language models (LLMs) to specific tasks or user preferences, a process called LLM adaptation, aims to improve performance and efficiency for diverse applications. Current research focuses on parameter-efficient fine-tuning techniques, such as low-rank adaptation and methods employing mixtures of experts or attention head modifications, to minimize computational costs and memory overhead while maintaining accuracy. These advancements are crucial for deploying LLMs on resource-constrained devices and for mitigating risks associated with adapting models using potentially malicious data. The resulting improvements in efficiency and controllability are significant for both scientific understanding of LLMs and their practical deployment across various industries.

Papers

December 9, 2023

Aligner: One Global Token is Worth Millions of Parameters When Aligning Large Language Models
Zhou Ziheng, Yingnian Wu, Song-Chun Zhu, Demetri Terzopoulos
Parameter Efficient Fine Tuning Many Parameter Value Alignment LLM Adaptation Worth Multiple Word Aligner Model

November 17, 2023

Camels in a Changing Climate: Enhancing LM Adaptation with Tulu 2
Hamish Ivison, Yizhong Wang, Valentina Pyatkin, Nathan Lambert, Matthew Peters, Pradeep Dasigi, Joel Jang, David Wadden, Noah A. Smith, Iz Beltagy, Hannaneh Hajishirzi
Large Language Model Language Model Instruction Tuning Electric SHEEP LLM Adaptation Climate Domain

October 17, 2023

Last One Standing: A Comparative Analysis of Security and Privacy of Soft Prompt Tuning, LoRA, and In-Context Learning
Rui Wen, Tianhao Wang, Michael Backes, Yang Zhang, Ahmed Salem
Context Learning Privacy Policy Security Related Private Data Privacy Analysis LLM Adaptation Soft Prompt Tuning Model Stealing

October 4, 2023

Sweeping Heterogeneity with Smart MoPs: Mixture of Prompts for LLM Task Adaptation
Chen Dun, Mirian Hipolito Garcia, Guoqing Zheng, Ahmed Hassan Awadallah, Anastasios Kyrillidis, Robert Sim
Mixture Component Complex Prompt Prompt Tuning Task Specification LLM Adaptation Heterogeneous Task Cleaning Robot Unknown Heterogeneity

June 8, 2023

An adaptive augmented Lagrangian method for training physics and equality constrained artificial neural networks
Shamsulhaq Basir, Inanc Senocak
Neural Network Constrained Optimization Theoretical Physic Navier Stokes LLM Adaptation Lagrangian Method Substantive Equality Incompressible Fluid

June 4, 2023

OWQ: Outlier-Aware Weight Quantization for Efficient Fine-Tuning and Inference of Large Language Models
Changhun Lee, Jungyu Jin, Taesu Kim, Hyungjun Kim, Eunhyeok Park
Large Language Model Scientific Inference Balancing Weight LLM Adaptation Low Precision Representation

LLM Adaptation

Papers

Aligner: One Global Token is Worth Millions of Parameters When Aligning Large Language Models

Camels in a Changing Climate: Enhancing LM Adaptation with Tulu 2

Last One Standing: A Comparative Analysis of Security and Privacy of Soft Prompt Tuning, LoRA, and In-Context Learning

Sweeping Heterogeneity with Smart MoPs: Mixture of Prompts for LLM Task Adaptation

An adaptive augmented Lagrangian method for training physics and equality constrained artificial neural networks

OWQ: Outlier-Aware Weight Quantization for Efficient Fine-Tuning and Inference of Large Language Models