Large language models implicitly learn to straighten neural sentence trajectories to construct a predictive representation of natural language

📄 arXiv: 2311.04930v1 📥 PDF

作者: Eghbal A. Hosseini, Evelina Fedorenko

分类: cs.CL, cs.AI

发布日期: 2023-11-05

备注: 37th Conference on Neural Information Processing Systems (NeurIPS 2023). 20 pages, 5 main figures, 7 supplementary figures

💡 一句话要点

提出神经轨迹直线化方法以提升语言模型预测能力

🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)

关键词: 自回归模型 语言预测 神经网络 轨迹直线化 变换器模型 几何特征 自然语言处理

📋 核心要点

  1. 现有的自回归变换器模型在语言预测任务中表现良好，但其内部表示的几何特征尚未得到充分理解。
  2. 论文提出通过量化句子的神经轨迹直线度，探讨如何通过直线化轨迹来提升语言模型的预测能力。
  3. 实验结果表明，经过训练的模型在深层网络中轨迹的曲率逐渐减小，且更好的预测能力与轨迹直线化程度相关。

📝 摘要（中文）

预测即将发生的事件对我们与环境的互动至关重要。经过训练的变换器模型在下一个单词预测任务中，似乎构建了能够支持多种下游任务的语言输入表示。本文提出了一种假设，探讨自回归变换器的预测表示如何受到预测目标的影响。我们量化了句子的神经轨迹直线度，并通过四个主要发现支持轨迹直线化假设，表明更直的轨迹有助于通过线性外推进行预测。这些结果为自回归模型内部表示的几何特征如何支持下一个单词预测提供了可能的机制。

🔬 方法详解

问题定义：本文旨在探讨自回归变换器模型在语言预测中的内部表示几何特征，现有方法未能充分揭示其对预测能力的影响。

核心思路：通过量化句子的神经轨迹直线度，提出轨迹直线化假设，认为更直的轨迹有助于提高预测性能。

技术框架：研究采用1维曲率度量来量化句子的直线度，分析不同层次的模型表现，并通过对比生成序列与真实语料的曲率进行验证。

关键创新：提出了轨迹直线化假设，揭示了自回归模型内部表示几何特征与预测能力之间的关系，这一视角在现有研究中较为新颖。

关键设计：使用1维曲率度量来量化直线度，分析模型在不同层次的曲率变化，结合下一个单词预测的表现进行综合评估。实验结果显示，未训练模型不具备相同的轨迹直线化特征。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

实验结果表明，经过训练的模型在深层网络中曲率显著降低，且在下一个单词预测任务中表现更佳。具体而言，曲率的减少与模型的预测能力呈正相关，生成的序列曲率低于真实语料的延续，显示出模型偏好直线化轨迹。

🎯 应用场景

该研究的潜在应用领域包括自然语言处理、对话系统和文本生成等。通过提升语言模型的预测能力，可以改善人机交互的流畅性和准确性，进而推动智能助手和自动翻译等技术的发展。

📄 摘要（原文）

Predicting upcoming events is critical to our ability to interact with our environment. Transformer models, trained on next-word prediction, appear to construct representations of linguistic input that can support diverse downstream tasks. But how does a predictive objective shape such representations? Inspired by recent work in vision (Henaff et al., 2019), we test a hypothesis about predictive representations of autoregressive transformers. In particular, we test whether the neural trajectory of a sentence becomes progressively straighter as it passes through the network layers. The key insight is that straighter trajectories should facilitate prediction via linear extrapolation. We quantify straightness using a 1-dimensional curvature metric, and present four findings in support of the trajectory straightening hypothesis: i) In trained models, the curvature decreases from the early to the deeper layers of the network. ii) Models that perform better on the next-word prediction objective exhibit greater decreases in curvature, suggesting that this improved ability to straighten sentence trajectories may be the driver of better language modeling performance. iii) Given the same linguistic context, the sequences that are generated by the model have lower curvature than the actual continuations observed in a language corpus, suggesting that the model favors straighter trajectories for making predictions. iv) A consistent relationship holds between the average curvature and the average surprisal of sentences in the deep model layers, such that sentences with straighter trajectories also have lower surprisal. Importantly, untrained models do not exhibit these behaviors. In tandem, these results support the trajectory straightening hypothesis and provide a possible mechanism for how the geometry of the internal representations of autoregressive models supports next word prediction.