Towards Building a Foundation AI Model for Railway Safety

Papalexakis, Evangelos; Chen, Jia; Dong, Yue

i

Towards Building a Foundation AI Model for Railway Safety

2025-09-30
By Papalexakis, Evangelos ; Chen, Jia ; Dong, Yue

File Language:

English

Details

Creators:

Papalexakis, Evangelos ; Chen, Jia ; Dong, Yue
Corporate Creators:

University Transportation Center for Railway Safety (UTCRS). University of California Riverside (UCR)
Corporate Contributors:

United States. Department of Transportation. Office of the Assistant Secretary for Research and Technology ; United States. Department of Transportation. University Transportation Centers (UTC) Program
Subject/TRT Terms:
Publication/ Report Number:

UTCRS-UCR-O6CY24
Resource Type:

Tech Report
Geographical Coverage:

United States
Edition:

June 1, 2024 - August 31, 2025
Corporate Publisher:

University Transportation Center for Railway Safety (UTCRS) Tier-1 University Transportation Center (UTC)
Abstract:

Large language models (often also termed foundation models) are revolutionizing multiple aspects of every-day life and work, with the use of artificial intelligence (AI) agents like ChatGPT being now commonplace and transforming life, work, and scientific discovery as we know it. In this project, we harness the power of such foundation models in order to transform railway safety. Specifically, we build proof-of-concept prototypes that demonstrate the viability of foundation models for railway safety question-answer generation and accident report-news article alignment. Furthermore, we develop algorithms for the improvement of performance of large language models, including hallucination detection foundation and model denoising.
Format:

PDF
Funding:

69A3552348340
Download URL:

https://rosap.ntl.bts.gov/view/dot/88689/dot_88689_DS1.pdf
File Type:

[PDF - 1.44 MB]
Collection(s):

Artificial Intelligence Deployment Center of Excellence (AI COE) ; University Transportation Centers
Main Document Checksum:

urn:sha-512:f725005ee72d18ab179032fcd365a2c1a16443d6fd2e1fe7bbd5ff024612b6ce8bb35a6844b34d8fda6f561239c702a4226cff23932ca4a85bc1467361f8b415

File Language:

English

ON THIS PAGE

Details

Found an issue?

Send us an email at:

publishinghd@cdc.gov

ROSA P serves as an archival repository of USDOT-published products including scientific findings, journal articles, guidelines, recommendations, or other information authored or co-authored by USDOT or funded partners. As a repository, ROSA P retains documents in their original published format to ensure public access to scientific information.