arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2605.21726cs.CLcs.AI
收藏

基于概率的大型语言模型归因

Probabilistic Attribution For Large Language Models

Argonne Leadership Computing Facility(阿贡领导计算设施);Argonne National Laboratory(阿贡国家实验室);Mathematics and Computer Science Division(数学与计算机科学 division);Department of Computer Science(计算机科学系)

Shilpika Shilpika, Carlo Graziani, Bethany Lusch, Venkatram Vishwanath, Michael E. Papka

AI总结:

本文提出了一种模型无关的概率性token归因度量,通过贝叶斯法则反向计算下一个token的对数概率,以捕捉模型对token序列分布的内部表示,从而提高大型语言模型的可解释性。

摘要:

大型语言模型(LLMs)生成性的特性体现在它们计算每个响应token的条件概率,以根据先前的token进行采样。这些概率编码了模型在训练中学习的分布结构,并在推理中加以利用。在本文中,我们利用这些概率将LLMs置于随机过程的数学理论框架中。我们使用此框架设计了一种模型无关的概率性token归因度量,通过贝叶斯法则反向计算下一个token的对数概率,以捕捉模型对token序列分布的内部表示。该表示独立于模型的计算结构。此表示给出了响应给提示的条件概率,以及在移除一个token后的响应给提示的条件概率。我们的归因分数是这两个概率比值的对数。我们进一步计算了单个提示token分布的熵,条件于剩余的上下文。熵与归因分数之间的相互作用揭示了LLM的行为。我们评估了8个模型在7个提示上的表现,并调查了异常、token敏感性、响应稳定性、模型稳定性以及训练收敛性,从而提高了可解释性,并引导用户关注生成中不确定或不稳定的部分。

英文摘要:

The generative nature of Large Language Models (LLMs) is reflected in the conditional probabilities they compute to sample each response token given the previous tokens. These probabilities encode the distributional structure that the model learns in training and exploits in inference. In this work, we use these probabilities to situate LLMs within the mathematical theory of stochastic processes. We use this framework to design a model-agnostic probabilistic token attribution measure, using Bayes rule to invert the next-token log-probabilities so as to capture the models internal representation of the distribution over token sequences. The representation is independent of the models computational structure. This representation yields the conditional probability of the response given the prompt, and of the response given the prompt with a token marginalized away. Our attribution score is the log of the ratio of these probabilities. We further compute the entropies of a single prompts token distributions, conditioned on the remaining context. The interplay between entropy and attribution score sheds light on LLM behavior. We evaluate 8 models across 7 prompts and investigate anomalies, token sensitivity, response stability, model stability, and training convergence, thereby improving interpretability and guiding users to focus on uncertain or unstable parts of the generation.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑