🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

How is the portrayal of violence in media changing over time? CSSLab at Penn researcher Amir Tohidi and team analyzed over 166K films using natural language processing, finding a rise in violent language. Details: bit.ly/4jcvy2y The Wharton School Annenberg School Penn Engineering

Join us in Philadelphia from June 30 to July 11, 2025!

In their new PNASNews paper, Rob Kuan, co-director Katy Milkman, & collaborators found that reminder emails can reduce missed payments among federal student loan borrowers. Read more about these scalable, low-cost nudges here: bit.ly/3WU2iUG

📣New Paper in PNAS: Short-term exposure to filter-bubble recommendation systems has limited polarization effects: Naturalistic experiments on YouTube pnas.org/doi/10.1073/pn…

Want to learn about computational social science *for free* and launch interdisciplinary research projects? We are so excited to announce there will be *26* Summer Institutes in Computational Social Science this year! Apply to one of them here: sicss.io/locations

I recently contributed a piece to Nature Astronomy on how large language models might reshape astronomy researcher identity and education - a topic I've been thinking about quite a bit lately. The piece is available here: rdcu.be/ebFws

One more week until applications are due. Don't miss your chance to attend SICSS-Penn this summer!

#ICLR2025 Thrilled to share that our ICLR 2026 paper has been accepted for an oral presentation! We show that LLMs represent political ideologies on a linear axis—and that we can detect and even steer their ideological perspectives. In our paper: - We find a linear "map" of

We test the relationship b/w multiple historical cultural patterns & modern-day language using #NLProc across two countries (China & Japan) & two social media platforms (Weibo & Twitter). Paper out in Humanities & Social Sciences Communications! nature.com/articles/s4159… Summary thread by Thomas Talhelm 🔽

The application deadline for SICSS-Penn is extended for one week. Submit your application by March 14 to attend our amazing program!

In this Perspective, Dörr et al. argue that research on social media has mainly focused on anti-social behaviours and call for more research on the ways in which social media platforms can empower pro-social behaviour. nature.com/articles/s4156…

Excited to share our latest research, published in PNAS Nexus! 🌍 w/ Stefano Balietti, Sam Fraiberger 🔎🌍 , anca.balietti 📄academic.oup.com/pnasnexus/adva… #ClimateChange #ScienceCommunication #ClimateAction

Paper day: Led by the one and only Tomasz Różański, we conducted a painstakingly detailed study on scaling spectral emulations with Transformers. We learned how to optimally predict the number of training samples needed in balance with model size and hyperparameter choices. The

Check out our latest paper! We test different strategies and models for prompting and fine tuning LLMs to perform stance classification tasks. dl.acm.org/doi/10.1145/37…

Meet Ai2 Paper Finder, an LLM-powered literature search system. Searching for relevant work is a multi-step process that requires iteration. Paper Finder mimics this workflow — and helps researchers find more papers than ever 🔍

📢 Our paper "AI safety for everyone" is out at Nature Machine Intelligence Nature Machine Intelligence . We challenge the narrative that AI safety is primarily about minimizing existential risks from AI. Why does this matter? A 🧵

Mixed feelings about silicon subjects (LLM predictions of human behavior) as replacements for human subjects? Consider the mixed subjects design. 🚨Now published at Sociological Methods and Research🚨 doi.org/10.1177/004912…

Updated version now out Sociological Methods & Research Paper link: journals.sagepub.com/doi/10.1177/00… Preprint link: osf.io/preprints/soca…

How can the community create high-quality trillion token scientific datasets from literature for training scientific foundation models? Meet AdaParse: A new adaptive PDF parsing engine that delivers 17x throughput while maintaining extraction accuracy by intelligently matching

AstroSage 70B (reasoning) - open weight - is here! arxiv.org/abs/2505.17592 huggingface.co/AstroMLab/Astr… Key takeaways: - AstroSage 70B performs best on AstroBench (Astronomy Q&A), outperforming all models (o3, Claude-3.7-Sonnet, Claude-4-Sonnet, Gemini-2.5-Pro) - CPT + SFT on