Stefano Sangiovanni

PhDing at @nasp_school @LaStatale. Currently visiting @tcddublin
Computational methods, Valence Politics. Just a nerd with some muddling through skills.

[Stefano Sangiovanni @SteSangio]

goat of open source

[antirez @antirez]

Welcome to DS4, a specialized inference engine for DeepSeek v4 Flash. https://t.co/UrUJz5I2R1

This project would have been impossible without the existence of llama.cpp and GGML and the work of @ggerganov and all the other contributors. Thanks!

[Mihai Simion @faustocoppi60]

This new emoji sums up today's Milano - Sanremo. 🫪
#MilanoSanremo

[World of Statistics @stats_feed]

🇮🇪 After 2 months of rain, today Ireland is expecting its first rain free day of 2026.

[haeju.eth @JeongHaeju]

“I just vibe coded a Polymarket insider trading detection tool with Claude code”

[FischerKing @FischerKing64]

29 years old, Stanford degree, chess grandmaster, successful teacher, commentator and streamer.

He’s been under intense pressure because of false cheating allegations, and showed signs of strain on stream. Good to reach out to friends when you know they’re having a tough time.

[International Chess Federation @FIDE_chess]

GM Daniel Naroditsky passed away. He was a talented chess player, commentator, and educator. FIDE extends its deepest condolences to Daniel’s family and loved ones.

[Lukáš Ronald Lukács @lucxsronald]

PASSION ❤️

📷: colnagoworld #ILombardia

[Lukáš Ronald Lukács @lucxsronald]

TADEJ POGAČAR IS CHAMPION OF THE WORLD ❤️

Remco fought hard, but still nice second place. Healy amazing bronze. #Kigali2025

[Luc Grefte @LucGrefte]

If you ever feel bad just remember the generation that had to race against Tadej Pogačar

[SESP @sesp_j]

📢NEW ARTICLE!📢 🇮🇹
Coalition formation, policy compromises and blame sharing: the strategic use of appeals for responsibility in Italian investiture debates

Read the article by @AndreaCeron83 @AlessioScopell5 & @GrrchanTamara here 👇
https://t.co/dfdvsREBWm

[Kobi Hackenburg @KobiHackenburg]

Today (w/ @UniofOxford @Stanford @MIT @LSEnews) we’re sharing the results of the largest AI persuasion experiments to date: 76k participants, 19  LLMs, 707 political issues.

We examine “levers” of AI persuasion: model scale, post-training, prompting, personalization, & more
🧵

[Alice Malmberg @alicemalmberg1]

My new R package with @SharifAmlani is live on Github! Introducing cjointplots, beautiful AMCE and marginal mean plotting for conjoint results: https://t.co/99rFSB7Iku

[Ronde van Vlaanderen @RondeVlaanderen]

Aura. 🫡 #RVV25 #FLCS

[λux @novasarc01]

this is really a brilliant work by team anthropic in interpretability research…understanding the fundamental nature of LLMs and how they actually work internally is a huge task for researchers…the new research actually explains it through circuit tracing and attribution graphs

[Kaggle @kaggle]

📢 The Kaggle GenAI Intensive with Google returns!

Taking place on Mar 31-Apr 4, this no-cost course is designed to help you gain a deeper understanding of the fundamental technologies and techniques behind Generative AI through theory, hands-on learning and community engagement.

More info and registration: https://t.co/YW02YAOJzu

[Neil Renic @NC_Renic]

anonymous reviewers recommending their own work

[andrea pedrazzani @andrea_piter]

Upcoming seminar:
Analyzing Public Perceptions of Government Actions & Political Scandals Through Survey Experimental Designs,
POLITICAL INSTITUTIONS & REGIMES SEMINAR SERIES #PIRSS🔝
18.12.24, 10:30, Conservatorio seminar room
@umbertoplatini @SteSangio A.Cassani @GrrchanTamara

[Stefano Sangiovanni @SteSangio]

@joon_s_pk @Carboaceto !! :O

[Mike Burnham @ML_Burn]

New Pre-print out today! We're releasing Political DEBATE -- a new set of language models for zero/few-shot classification of political text. The models are open source, small enough to run on your laptop, and as good as proprietary LLMs within domain. https://t.co/4Q0bVOgMgf

[Martin JJ. Bucher @mnbucher]

In our latest paper, we challenge the claim that *zero-shot* generative AI models outperform smaller LLMs for text classification — and empirically show that ‘small’ LLMs fine-tuned on a few hundred samples consistently beat larger, zero-shot models across diverse tasks. (1/7)