Many thanks to all the coauthors involved in the project @JoeunYk05 @psyonp @keenansamway @bschoelkopf @albecazzaniga @radamihalcea @ZhijingJin
English
francesco ortu
34 posts
francesco ortu
@francescortu
NLP & Interpretability | PhD Student @UniTrieste & Data Engineering Lab @AreaSciencePark | Prev @MPI_IS intern https://t.co/RM03tgnbJP
Trieste, Friuli-Venezia Giulia Katılım Mart 2011
1.2K Takip Edilen267 Takipçiler
I'm really excited to be in Paris for the second annual conference of the @IASEAIorg, where yesterday I presented our newly released paper: “Preserving Historical Truth: Detecting Historical Revisionism in Large Language Models” 🇫🇷
Paper link below ⬇️
English
francesco ortu retweetledi
First day at UNESCO: We presented our Detecting LLM Historical Revisionism paper by @FrancescOrtu @JoeunYk05 @psyonp @KeenanSamway @BSchoelkopf @AlbeCazzaniga @RadaMihalcea @ZhijingJin and will present Accidental Vulnerability by @psyonp @SimkoSamuel @KellinPelrine @ZhijingJin!
Zhijing Jin@ZhijingJin
Excited to have 3 accepted papers & 9 members of our @JinesisLab at #IASEAI2026, held at UNESCO, Paris🇫🇷! We reveal hidden authoritarian biases in #LLMs, and that fine-tuning can quietly erode model safety, exploring the risks we don't always see in AI 🔍🛡️ 🧵👇
English
@benno_krojer Thanks for sharing! We were honestly pretty surprised when we first saw the narrow gate effect and how strongly it shows up in the models
English
A paper that should get more attention, for those interested in building truly multimodal models (aka not just plugging stuff post hoc into LLMs):
arxiv.org/abs/2412.06646
TLDR: Counterintuitively, native multimodal modes seem *less* unified internally (interp)
English
francesco ortu retweetledi
We're at #NeurIPS2025 with papers, posters, and talks across many workshops. Come learn about our latest research and explore our newest breakthroughs in #LLMs, #Causality, #AIforScience, and many others!
English
francesco ortu retweetledi
Excited to present "The Narrow Gate: Localized Image-Text Communication in Native Multimodal Models" at #NeurIPS2025 and #EurIPS2025!
📍NeurIPS: Poster Session 6 #1114 (Fri Dec 5 4pm-7pm)
📍EurIPS: Poster Session D3 #88 (Wed Dec 3 10:30am-12:30pm)
🧵👇
English
francesco ortu retweetledi
Excited to share that 2/2 papers from our Lab (LADE) were accepted to #NeurIPS2025 (one spotlight 🎉)
Great work from all the students and collaborators involved!
@AreaSciencePark @aleserra1998 @lorebasile @francescortu @lucrevaleriani @DiegoDoimo @ansuin @FrancescoLocat8
English
Thrilled to announce that our paper is accepted at #NeurIPS 2025!! See you in San Diego! 🇺🇸
francesco ortu@francescortu
🚨🚨 Excited to share our latest paper, now on @arxiv! 🖼️ We studied how unified VLMs, trained to generate both text and images (e.g., @MetaAI's Chameleon), exchange information between modalities, comparing them to standard VLMs. Deep dive:👇
English
Check out our follow-up paper on the competition between visual context and inner knowledge in VLMs! I'm very happy about this work :) Link in the thread 🧵
Zhijing Jin@ZhijingJin
Our "Competitions of Mechanisms" paper proposes an interesting way to interpret LLM behaviors thru how it handles multiple conflicting mechanisms. E.G., in-context knowledge vs. in-weights knowledge🧐This is an elegant philophical way of thinking --
English
francesco ortu retweetledi
🌍 How do #LLMs handle trolley problems across cultures? We test them with 98K dilemmas in 107 languages, grounded in 40M+ human moral judgments.
💡 Spotlight @ICLR2025 in Singapore✈️| Best Paper @pluralistic_ai Workshop #NeurIPS2024
📄 Paper: arxiv.org/abs/2407.02273 🧵👇
English
Really excited to see this work recognized! I'm so thankful to have been involved alongside such brilliant minds 🎉
Zhijing Jin@ZhijingJin
Honored to receive the Best Paper Award (the top prize!) at the #NeurIPS2024 Workshop on Pluralistic Alignment @pluralistic_ai! Many thx to my wonderful coauthors, who taught me so much about this interdisciplinary field of #LLMs and moral reasoning: @maxhkw @giorgiopiatti @sydneymlevine @Jiarui_Liu_ @fer_adauto @francescortu Andras Strausz @mrinmayasachan @radamihalcea @YejinChoinka @bschoelkopf. Also thank you so much to all the co-organizers and especially @Ruyuan_Wan for the wonderful photo capturing this memorable moment! 🎉 Check out our "Language Model Alignment in Multilingual Trolley Problems" at arxiv.org/pdf/2407.02273!
English
Thanks, @DiegoDoimo and @albecazzaniga, for the fantastic mentorship and support! 🙏🎉 They are also attending #NeurIPS, so feel free to reach out to them to discuss our results. I’m excited to keep pushing forward on these topics! 🚀
English
It was super fun to take our first step in interpreting multimodal LLMs, working closely with the brilliant @aleserra1998 and @PanizonEmanuele , alongside the amazing team at LADE @AreaSciencePark : @lucrevaleriani @lorebasile @ansuin @DiegoDoimo @albecazzaniga
English
