iiinova — where your digital self keeps evolving

Peter Carragher

@pcarragher.bsky.social

CS PhD student at CMU Retrieval, Webgraphs, Source Credibility, Adversarial Adaptation

Peter Carragher @pcarragher.bsky.social · 07/08/2025

Last week, I presented my recent paper on memorization and hallucination in LVLMs at @l2m2workshop.bsky.social #ACL2025. TLDR: model memorization may be more prevalent over image sources vs. text! Details here: aclanthology.org/2025.l2m2-1....

140

Peter Carragher @pcarragher.bsky.social · 24/06/2025

Excited to have two workshop papers and one main conference paper that I've been involved in being presented at @icwsm.bsky.social! Thanks @kingcatherine.bsky.social and @evanup.bsky.social for letting me tag along. Details below.

193

Peter Carragher @pcarragher.bsky.social · 15/06/2025

Finetuning helps identify counterfactual samples and prevent hallucinations.

100

Peter Carragher @pcarragher.bsky.social · 15/06/2025

Preprint! Vision-language models like GPT-4o fail to identify multimodal knowledge conflicts and frequently hallucinate on counterfactual samples. Try out the batsman example below (promt = "what is he holding?").

110