Research and Analysis of Large Language Model Synthetic Data Generation and Bias and Illusion Problems

DOI

10.2991/978-94-6239-648-7_73How to use a DOI?

Keywords

Computer vision; Natural language process; Reinforcement learning; Logical illusion recognition

Abstract

This dissertation examines the relevance of large language models (LLMs) in enhancing time-series data applications, e.g., climate forecasting, traffic control, and finance, in which quality data is critically important in the prediction and making of decisions. The importance of overcoming the limitations of LLM in these aspects is to reduce real-life problems, such as data shortage, related privacy issues, and elevated labeling expenses to make the model reliable and allow it to be used in industries where time-related data are needed. This dissertation presents the key problems with big language models (LLM) in time-Series scenarios with a focus on the joint optimization problem that entails synthetic data generation, bias mitigation, and illusion detection. Meeting three key issues of the current technologies, such as the lack of domain adaptability, the inability to remove multidimensional biases and detect sophisticated logical illusions, the study offers a technical solution based on the specifics of time-series data. The efficiency of the given solution is confirmed with the help of the real research that is performed on the PyTorch experimental platform based on the data of the Daily Climate Forecasting, traffic prediction, and stock sentiment analysis. Lastly, a three phase technical scheme of explicit temporal semantic modeling + cross-view consistency constraints + domain knowledge fusion is built, which offers Conceptual foundation and pragmatic directions to an enhanced trustworthiness of the LLMs in the time-series area.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Bolin Zhang
PY  - 2026
DA  - 2026/04/24
TI  - Research and Analysis of Large Language Model Synthetic Data Generation and Bias and Illusion Problems
BT  - Proceedings of the International Workshop on Advances in Deep Learning for Image Analysis and Computer Vision (IWADIC 2025)
PB  - Atlantis Press
SP  - 676
EP  - 684
SN  - 2352-538X
UR  - https://doi.org/10.2991/978-94-6239-648-7_73
DO  - 10.2991/978-94-6239-648-7_73
ID  - Zhang2026
ER  -