Semantic topologies in the recursive application of generative AI models

Ben Swift; Sungyeon Hong

doi:10.1109/SMC58881.2025.11342470

Semantic topologies in the recursive application of generative AI models

Ben Swift, Sungyeon Hong

School of Cybernetics

Research output: Contribution to conference › Paper › peer-review

Abstract

Text-to-image and image-to-text models allow automated (but imperfect) semantic translation across modalities. This paper presents results and preliminary analysis of an empirical study of recursive information processing in popular open-weight generative artificial intelligence (genAI) models such as FluxSchnell and BLIP-2. Through clustering and topological data analysis we show some of the ways that different genAI models and initial prompts give rise to different semantic embedding trajectories, and suggest some ways forward for understanding how semantic information is transmitted through these types of complex information-processing systems.

Original language	English
Pages	664-667
Number of pages	4
DOIs	https://doi.org/10.1109/SMC58881.2025.11342470
Publication status	Published - 2025
Event	2025 IEEE International Conference on Systems, Man, and Cybernetics (SMC) - Vienna, Austria Duration: 5 Oct 2025 → 8 Oct 2025 https://ieeexplore.ieee.org/xpl/conhome/11342430/proceeding

Conference

Conference	2025 IEEE International Conference on Systems, Man, and Cybernetics (SMC)
Abbreviated title	SMC
Country/Territory	Austria
City	Vienna
Period	5/10/25 → 8/10/25
Internet address	https://ieeexplore.ieee.org/xpl/conhome/11342430/proceeding

Access to Document

10.1109/SMC58881.2025.11342470

Cite this

@conference{f56ee73db53d4c3996cec16b7ef8fa90,

title = "Semantic topologies in the recursive application of generative AI models",

abstract = "Text-to-image and image-to-text models allow automated (but imperfect) semantic translation across modalities. This paper presents results and preliminary analysis of an empirical study of recursive information processing in popular open-weight generative artificial intelligence (genAI) models such as FluxSchnell and BLIP-2. Through clustering and topological data analysis we show some of the ways that different genAI models and initial prompts give rise to different semantic embedding trajectories, and suggest some ways forward for understanding how semantic information is transmitted through these types of complex information-processing systems.",

author = "Ben Swift and Sungyeon Hong",

year = "2025",

doi = "10.1109/SMC58881.2025.11342470",

language = "English",

pages = "664--667",

note = "2025 IEEE International Conference on Systems, Man, and Cybernetics (SMC), SMC ; Conference date: 05-10-2025 Through 08-10-2025",

url = "https://ieeexplore.ieee.org/xpl/conhome/11342430/proceeding",

}

TY - CONF

T1 - Semantic topologies in the recursive application of generative AI models

AU - Swift, Ben

AU - Hong, Sungyeon

PY - 2025

Y1 - 2025

N2 - Text-to-image and image-to-text models allow automated (but imperfect) semantic translation across modalities. This paper presents results and preliminary analysis of an empirical study of recursive information processing in popular open-weight generative artificial intelligence (genAI) models such as FluxSchnell and BLIP-2. Through clustering and topological data analysis we show some of the ways that different genAI models and initial prompts give rise to different semantic embedding trajectories, and suggest some ways forward for understanding how semantic information is transmitted through these types of complex information-processing systems.

AB - Text-to-image and image-to-text models allow automated (but imperfect) semantic translation across modalities. This paper presents results and preliminary analysis of an empirical study of recursive information processing in popular open-weight generative artificial intelligence (genAI) models such as FluxSchnell and BLIP-2. Through clustering and topological data analysis we show some of the ways that different genAI models and initial prompts give rise to different semantic embedding trajectories, and suggest some ways forward for understanding how semantic information is transmitted through these types of complex information-processing systems.

UR - https://github.com/ANUcybernetics/panic-tda/

U2 - 10.1109/SMC58881.2025.11342470

DO - 10.1109/SMC58881.2025.11342470

M3 - Paper

SP - 664

EP - 667

T2 - 2025 IEEE International Conference on Systems, Man, and Cybernetics (SMC)

Y2 - 5 October 2025 through 8 October 2025

ER -

Semantic topologies in the recursive application of generative AI models

Abstract

Conference

Access to Document

Other files and links

Fingerprint

Cite this