Tuğrulcan Elmas 🇹 🇯

How to tell if a trending slogan is a coordinated campaign? We present a new campaign dataset & detection method in Large Engagement Networks for Classifying Coordinated Campaigns and Organic Twitter Trends @ICWSM & Density-aware Walks for Coordinated Campaign Detection @ECMLPKDD

In the US in this week alone we've got a ramp up of authoritarian consolidation and cultural engineering. Everywhere from universities to sports teams I draw attention to entertainment media spaces in which those processes happen. Like cooking shows. buff.ly/5bezcnG

“We must ensure that free speech flourishes in the era of AI and that AI procured by the Federal government objectively reflects truth rather than social engineering agendas…. eliminate references to misinformation, Diversity, Equity, and Inclusion, and climate change.” Hmmmmm

holy claude, that's cool!

I asked Gemini how squid game should have ended. It gaslighted me by saying there is no Squid Game season 3, then described "the fictional final" which is exactly the same as the real final weird hallucination🧐

Nice AI drama

yOu Do Nt nEEd a PhD tbe an AI rschr :S:S:S

"How Many Instructions Can LLMs Follow at Once?" In this paper they found that leading LLMs can satisfy only about 68% of 500 concurrent instructions, showing a bias toward earlier instructions.

I didn't want to post on Grok safety since I work at a competitor, but it's not about competition. I appreciate the scientists and engineers at @xai but the way safety was handled is completely irresponsible. Thread below.

Tokenization has been the final barrier to truly end-to-end language models. We developed the H-Net: a hierarchical network that replaces tokenization with a dynamic chunking process directly inside the model, automatically discovering and operating over meaningful units of data

Dunno if Grok 4 is really that good but I'm wondering, was the Grok shootshow yesterday a viral campaign😂

Hey @PMinervini why can't you be like this

Non consensual deepfakes have always been a concern but they've been getting realllyy good with the recent LoRA methods This paper looks at a platform that hosts such finetuned LoRA models & shows how it disproportionately targets women (not just celebs) arxiv.org/abs/2505.03859

❓Why do the Nordics & Dutch speak English so much better than the Germans, Italians & French? ➡️New Working Paper: Out-of-School Learning: Subtitling vs. Dubbing and the Acquisition of Foreign-Language Skills w/ F. Baumeister & E. Hanushek nber.org/papers/w33984 A🧵1/12

An AI model (Llama 3.1 70B) fine-tuned on the results of 60,000 people in psychology experiments shows some real promise in using LLMs for studying human behavior. It predicts actual human behavior in held-out data & it generalizes to out-of-distribution tasks and experiments.

"in 2025 we will have flying cars" 😂😂😂

An interesting analysis of how cooking shows reinforce conservative gender stereotypes by creating conflicts that portray traditional women as heroes and those who deviate from these norms as villains.

So much going on, but I'm pleased to share my study of the vernacular politics of gender construction. As a former pastry chef, I gathered data from 600 hours of cooking shows to unpack what I term "conservative gender edutainment." Read @PoliticsGender cup.org/46pO43h

Based on recent events, Morgan J. Freeman's cryptic post likely refers to Soham Parekh, an Indian software engineer accused of working at multiple startups simultaneously without disclosure. On July 2, 2025, tech figures like Suhail Doshi reported firing Parekh for deceptive…