Using Laplace Transform To Optimize the Hallucination of Generation Models

Cheng Kang, Xinye Chen, Daniel Novak, Xujing Yao · 2026

Read Using Laplace Transform To Optimize the Hallucination of Generation Models by Cheng Kang, Xinye Chen, Daniel Novak, Xujing Yao (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-18022.txt.

Introduction

Kang and co-authors recast generation models as stochastic dynamical systems through control theory. They aim to curb confident errors, or hallucinations, in model output. A full Laplace solution is out of reach for such complex systems, they note. Still, simulated source responses track how training behaves in practice. That link guides a better optimization part, and Laplace analysis then trims hallucination at its root, they argue.

Catalog record

CREATORCheng Kang, Xinye Chen, Daniel Novak, Xujing Yao
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE31281
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.18022

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Using Laplace Transform To Optimize the Hallucination of Generation Models Cheng Kang, Xinye Chen, Daniel Novak, Xujing Yao Abstract To explore the feasibility of avoiding the confident error (or hallucination) of generation models (GMs), we formalise the system of GMs as a class of stochastic dynamical systems through the lens of control theory. Numerous factors can be attributed to the hallucination of the learning process of GMs, utilising knowledge of control theory allows us to analyse their system functions and system responses. Due to the high complexity of GMs when using various optimization methods, we cannot figure out their solution of Laplace transform, but from a macroscopic perspective, simulating the source response provides a virtual way to address the hallucination of GMs. We also find that the training progress is consistent with the corresponding system response, which offers us a useful way to develop a better optimization component. Finally, the hallucination problem of GMs is fundamentally optimized by using Laplace transform analysis. I INTRODUCTION Generation models, especially, large language models (LLMs) and large vision generation models (LVGMs), are suffering from the hallucination problem. Speaking of language generation and vision-language generation models, hallucination is referred to as a situation where the model generates content that is not based on factual or accurate information [ 24 ] . Hallucination in vision generation models is a phenomenon that models generate samples that lie completely out of the training distribution of the model [ 1 ] . One conversational study investigated the original of hallucination, and they found that hallucination is a prevalent issue in both dialog benchmarks and models [ 10 ] . Recent studies of diffusion generation models indicates that the hallucination phenomenon has a ”mode collapse” [ 31 ] which leads to a loss in diversity in the sampled distribution. So how does existing studies address hallucination? Focusing on the microscopic level, a large commonality in the majority of prior work (both in language generation models [ 26 , 9 ] and vision generation models [ 31 , 14 , 17 ] ) seeks to address hallucination by ameliorating the model. Retrieval augmented generation (RAG) improved the reliability of LLMs by reading extra knowledge from new data [ 26 ] . For language-vision models, existing studies [ 13 , 30 ] mitigated the issue of hallucination in image captioning models with a predominant focus on the “presence of objects”, specifically whether or not the objects depicted in a given image are accurately described by the text that is generated by the mode. Most advanced generation adversarial models (GANs) assembled some quality components [ 33 ] that wonderfully optimize the generation results. Paying attention to the local out-of-distribution (OOD) regions of images when using denoising diffusion probabilistic models (DDPM) to translate one image to another image, multiple local diffusion processes proved to reduce the hallucination of generated images [ 14 ] . However, to the best of our knowledge, apart from the problem of dataset, there is no research and discussion on the stability of these learning systems, as well as the fundamental reason why generation models have confident error or hallucination in general. Can we fundamentally figure out why generation models have hallucination from a macroscopic perspective? Numerous advanced components, especially, optimization methods [ 28 ] and loss functions [ 6 ] , have improved generation models, and their utilisation will affect the performance of GMs. Conducting a quantitative analysis on each of them can pave the way for their efficiency and development. In one node of the learning system, and in terms of analyzing a single component by computing their Laplace transform, their stability and convergence [ 21 ] can be evaluated to finally ensure that GMs are able to generate a reliable and stable output. How well does Laplace transform address the hallucination of GMs? Based on the Laplace transform analysis [ 21 ] , we find that using a proper optimizer (such as, using Adam optimizer on GAN [ 11 ] and DDPM [ 12 ] , using FuzzyPID optimizer [ 29 ] on CycleGAN [ 6 ] ), can benefit the learning process of GMs. The use of a proper component can ensure a harmony of the whole system, which can visually tell us which component should be used in which learning system. In this study, to guarantee the stability and convergence of GMs, we find that there are two ways that can address the hallucination of GMs via Laplace transform: (1) using Laplace transform to develop a proper component; (2) using Laplace transform to design a better learning system. Contributions are shown below. • Laplace transform analysis finds the fundamental reason why generation models have hallucination. • Optimization component should specifically designed to match the generation models; • The practical performance of every optimization component is highly consistent with their system response. II The Connection Between Control Theory and Learning Progress of GMs Figure 1 : Evolution process of the weight \theta({t}) for each component under the generation task. In Figure 1 , \theta_{max} is the maximum overshoot of training model, and t_{max} is the needed epoch or step when the training model reaches the strongest vibration. We assume that the overshoot and vibration is the source of hallucination. The optimal condition is that the training model has a proper maximum vibration (no vibration will make model learn slowly, for example, component B in Figure 1 ), a shorter climbing time t_{max} , a visual robustness, and a reliable and convergent result (the learning process will result in a long time to convergence or no convergence, for example, component A in Figure 1 ). Compared to the real GM itself, so as to obtain the optimal \theta^{\ast} , we should guarantee that the learning progress is associated with a stable training progress for a better result with a shorter training time. For a clear simulation analysis, we assume that the system function of every GM is a two-order system, and their system function can be given by 1/(s^{2}+P_{a}s+P_{b}) ), where P , P_{a} and P_{b} are the corresponding variables, respectively. III The Laplace Transform of Optimization Methods In this section, we review several widely used optimizers, such as SGD [ 25 , 7 , 32 ] , SGDM [ 23 , 19 ] , Adam [ 15 , 4 ] , PID-optimizer [ 28 ] and Gaussian LPF-SGD [ 3 ] . III-A SGD Optimization Method The parameter update rule of SGD from iteration t to t+1 is determined by \theta_{t+1}=\theta_{t}-r\frac{\partial L_{t}}{\partial\theta_{t}}, (1) where r is the learning rate. We now regard the gradient \partial L_{t}/\partial\theta_{t} as error e(t) in the PID control system [ 28 ] . Compared to the PID controller, we find that SGD can be viewed as one type of P controller with K_{p}=r . The system function of SGD finally becomes: \theta_{SGD}(s)=r. (2) III-B SGDM Optimization Method By leveraging historical gradients, SGDM trains a ANN more swiftly than SGD does. The rule of SGDM updating parameter is given by \left\{\begin{array}[]{l}V_{t+1}=\alpha V_{t}-r\partial L_{t}/\partial\theta_{t}\\ \theta_{t+1}=\theta_{t}+V_{t+1}\end{array}\right. (3) where V_{t} is a term that accumulates historical gradients. \alpha\in(0,1) is the factor that balances the past and current gradients. It is usually set to 0.9 .

Read the complete text