@iaugenstein.bsky.social
Professor at the University of Copenhagen. Explainable AI, Natural Language Processing, ML. Head of copenlu.bsky.social lab.
#NLProc #NLP #XAI
http://isabelleaugenstein.github.io/
@iaugenstein.bsky.social🚨 Last chance -- closing in two days: call for a postdoc position on mechanistic interpretability and AI reasoning in our @copenlu.bsky.social lab ⤵️
@aicentre.dk
#NLProc #XAI #academicjobs
@iaugenstein.bsky.social🚨Closing in two weeks: opportunity to apply for a postdoc position on mechanistic interpretability and AI reasoning in our @copenlu.bsky.social lab ⤵️
@aicentre.dk #NLProc #XAI #academicjobs
@siddheshp.bsky.socialReposted by @iaugenstein.bsky.socialPeople use LLMs to answer subjective, culturally situated questions. Most benchmarks ask whether the answer is correct, but they overlook the framing features (eg, insider positioning) 📄 New #EMNLP'26 paper: "Not What, But How: A Communicative Audit of LLM Response Framing"
@iaugenstein.bsky.socialOur paper on measuring epistemic diversity in LLMs is accepted to #EMNLP2026!
We find that diversity has improved, though LLM output is much less diverse than Web search, especially for non-English.
Paper: arxiv.org/pdf/2510.04226
Code/data: github.com/dwright37/ll...
#NLProc @copenlu.bsky.social
@sekh-copenlu.bsky.socialReposted by @iaugenstein.bsky.socialExcited to share that our work, "BiasGym: A Simple and Generalizable Framework for Analyzing and Removing Biases through Injection", was accepted at the EMNLP 2026 (Findings).
📄 Paper (Updated): arxiv.org/pdf/2508.08855
💻 Code: github.com/copenlu/Bias...
@sekh-copenlu.bsky.socialReposted by @iaugenstein.bsky.socialExcited to share that our work, "Multi-Step Knowledge Interaction Analysis via Rank-2 Subspace Disentanglement", was accepted at the EMNLP 2026 (Findings). Huge thanks to my amazing PhD supervisors: @iaugenstein.bsky.social, and @apepa.bsky.social! See you all in Budapest 🇭🇺!
@gretawarren.bsky.socialReposted by @iaugenstein.bsky.socialHow do people use images to spread misinformation online?🖼️
We developed a taxonomy, analysed 27k posts from X in 7 languages & found:
🔹Slanted framing of real images is more common than deepfakes and doctored images
🔹Vaccine misinfo borrows credibility via news screenshots
@iaugenstein.bsky.social📣 Hiring postdoc in Mechanistic Understanding of AI Reasoning 📣
⏳ Application deadline: 30 Sep 2026
🏁 Start date: 1 Feb 2027
💼 Contract length: 2 years
🏰 Workplace: @aicentre.dk @copenlu.bsky.social
ℹ️ Apply here: candidate.hr-manager.net/ApplicationI...
