arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2603.11228cs.CLcs.AIcs.LG
收藏

大语言模型中的马尔可夫生成链

Markovian Generation Chains in Large Language Models

ENS-PSL & CNRS-Lattice(ENS-PSL与CNRS-Lattice);MBZUAI & Ecole Polytechnique(MBZUAI与巴黎高等理工学院)

Mingmeng Geng, Amr Mohamed, Guokan Shang, Michalis Vazirgiannis, Thierry Poibeau

AI总结:

本文研究了大语言模型中迭代推理过程的演变,通过马尔可夫链建模发现句子多样性受温度参数和初始输入影响,揭示了迭代推理的动力学及其对多智能体系统的影响。

摘要:

大语言模型(LLMs)的广泛应用引发了一个重要问题:当文本被反复处理时,文本如何演变?本文将这种迭代推理过程定义为马尔可夫生成链,其中每一步都使用特定的提示模板和前一步的输出作为输入,不包含任何先前的记忆。在迭代改写和往返翻译实验中,输出要么收敛到一个较小的循环集,要么在有限的范围内继续生成新的句子。通过句子层面的马尔可夫链建模和对模拟数据的分析,我们表明,根据温度参数和初始输入句子等因素,迭代过程可以增加或减少句子多样性。这些结果为迭代LLM推理的动力学及其对多智能体LLM系统的影响提供了有价值的见解。

英文摘要:

The widespread use of large language models (LLMs) raises an important question: how do texts evolve when they are repeatedly processed by LLMs? In this paper, we define this iterative inference process as Markovian generation chains, where each step takes a specific prompt template and the previous output as input, without including any prior data. In iterative rephrasing and round-trip translation experiments, the output either converges to a small recurrent set or continues to produce novel sentences within a finite horizon. Through sentence-level Markov chain modeling and analysis of simulated data, we show that iterative process can either increase or reduce the diversity depending on factors such as the temperature parameter and the initial input sentence. These results offer valuable insights into the dynamics of iterative LLM inference and their implications for multi-agent LLM systems.

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。

↑