分离大模型的表征与预测，用多层隐状态提升问答准确率
InnerThoughts: Disentangling Representations and Predictions in Large Language Models
- 用多层隐藏状态构建独立预测模块，解耦表征与推理能力
- 在多个难题集上性能显著提升，接近微调效果
- 计算开销极小，适合资源受限场景使用
大型语言模型（LLMs）包含大量事实知识，通常通过多项选择题提示激发。模型在多个Transformer层中处理提示，逐步生成问题的隐藏表示，但最终仅使用最后一层和最后一个词元的隐藏状态进行答案预测。本文提出一种新方法：在训练数据上学习一个小型独立神经网络预测模块，输入为所有层在最后时间位置的隐藏状态，输出预测结果。该框架实现了模型表征能力与预测能力的解耦。在一系列困难基准测试中，该方法显著提升了性能，部分表现接近监督微调，但计算成本仅为后者的极小部分。
原文摘要 · Abstract (English)
Large language models (LLMs) contain substantial factual knowledge which is commonly elicited by multiple-choice question-answering prompts. Internally, such models process the prompt through multiple transformer layers, building varying representations of the problem within its hidden states. Ultimately, however, only the hidden state corresponding to the final layer and token position are used to predict the answer label. In this work, we propose instead to learn a small separate neural network predictor module on a collection of training questions, that take the hidden states from all the layers at the last temporal position as input and outputs predictions. In effect, such a framework disentangles the representational abilities of LLMs from their predictive abilities. On a collection of hard benchmarks, our method achieves considerable improvements in performance, sometimes comparable to supervised fine-tuning procedures, but at a fraction of the computational cost.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。