Large Language Model (LLM) conditioning refers to instructing an LLM to generate content in accordance with the norms and values of a specific culture, beliefs of a particular political orientation, or any desired text-specified semantic conditioning. Unfortunately, prompt engineering does not ensure that LLMs behave in accordance with a desired conditioning due to the inductive bias of the pre-training and alignment datasets. Prior works have focused on fine-tuning LLMs by directly conditioning the LoRA weights; however, such methods introduce a large number of parameters. As a remedy, we propose Zhyper, a parameter-efficient factorized hypernetwork framework that generates context-aware LoRA adapters from textual descriptions. Experiments on multiple benchmarks show that Zhyper achieves competitive performance with up to 26x fewer parameters than the state-of-the-art baselines. Furthermore, we extend Zhyper to cultural alignment, demonstrating improved generalization to out-of-domain settings and a better capturing of fine-grained contextual values.

翻译：大语言模型（LLM）条件化是指引导LLM按照特定文化的规范与价值观、特定政治立场的信念，或任何期望的文本指定语义条件来生成内容。遗憾的是，由于预训练和对齐数据集的归纳偏差，提示工程并不能确保LLM的行为符合期望的条件化。先前的研究集中于通过直接条件化LoRA权重来微调LLM，但这类方法会引入大量参数。作为改进，我们提出了Zhyper——一种参数高效的因子化超网络框架，能够从文本描述生成上下文感知的LoRA适配器。在多个基准测试上的实验表明，Zhyper以比最先进基线最多减少26倍的参数量，实现了具有竞争力的性能。此外，我们将Zhyper扩展至文化对齐任务，证明了其在领域外设置下具有更好的泛化能力，并能更精细地捕捉上下文价值。