Florian Eichin
@florian-eichin.com
PhD candidate at LMU Munich. Representations, model and data attribution, training dynamics. Strong opinions on coffee and tea ☕ https://florian-eichin.com
𝗚𝗿𝗮𝗱𝗶𝗲𝗻𝘁 𝗱𝗲𝘀𝗰𝗲𝗻𝘁 𝗶𝘀 𝗮𝗹𝗹 𝘆𝗼𝘂 𝗻𝗲𝗲𝗱 𝗳𝗼𝗿 𝗶𝗻𝘁𝗲𝗿𝗽𝗿𝗲𝘁𝗮𝗯𝗶𝗹𝗶𝘁𝘆? In our ICML paper, we show that it might just be. ExPLAIND is a method that unifies data attribution, model component attribution, and training dynamics by computing an exact decomposition of model behavior based on gradient products.
✨New paper✨ We find script (e.g. Cyrillic, Latin) to be a linear direction in the activation space of Whisper, enabling transliteration at test-time by adding such script directions to the activations — producing e.g. Cyrillic Japanese transcriptions.
Thank you to our great Munich Center for Machine Learning (@munichcenterml.bsky.social) for featuring me in this research film! Lots of great films clips with my MCML colleagues are available on MCML's YouTube channel. #ai #aiethics #mcml #philosophy youtu.be/KUqiY8o1yng?...
What is intelligence—and what kind of intelligence do we want in our future? With Prof. Sven Nyholm
YouTube video by MCML_Munich Center for Machine Learning
youtu.be
Munich Center for Machine Learning@munichcenterml.bsky.social · 12mo ago🎥 Who gets the credit, or the blame, when AI makes decisions? Sven Nyholm ( #LMU/ #MCML) reflects on how AI challenges our ideas of agency, credit, and blame — and why we need new ways of thinking about authorship, justice, and decision-making. www.youtube.com/watch?v=KUqi... #AI #Ethics
🕺🏼swing by our poster in Hall 4/5 on Wednesday, July 30 at 11:00 to chat with @florian-eichin.com and I to find out the answers to these questions 🛎️ bonus: to see the full poster 🫣🧩 #ACL2025 #NLProc
MaiNLP lab, LMU Munich@mainlp.bsky.social · last yr.📝Probing LLMs for Multilingual Discourse Generalization Through a Unified Label Set 🔎Do LLMs encode and generalize discourse knowledge across languages? 👥 @florian-eichin.com @janetlauyeung.bsky.social @mhedderich.bsky.social @barbaraplank.bsky.social 🔗 arxiv.org/abs/2503.10515 📁Main - Long
📝Probing LLMs for Multilingual Discourse Generalization Through a Unified Label Set 🔎Do LLMs encode and generalize discourse knowledge across languages? 👥 @florian-eichin.com @janetlauyeung.bsky.social @mhedderich.bsky.social @barbaraplank.bsky.social 🔗 arxiv.org/abs/2503.10515 📁Main - Long
Some recommendations for #ACL2025 👇 (join me and @janetlauyeung.bsky.social to talk about discourse generalization and probing!)
MaiNLP lab, LMU Munich@mainlp.bsky.social · last yr.Headed to ACL? MaiNLP & our most recent work will be there too👥📄 Come see what we’ve been working on!
Headed to ACL? MaiNLP & our most recent work will be there too👥📄 Come see what we’ve been working on!
XAI’s dogwater performance on the 2025 IMO confirms that their Grok 4 benchmark claims were hot air. Their eye popping metrics were down to the following innovations: - train on test - train on test - train on test
📄 [ACL 2025 main] LLMs instead of Human Judges? A Large Scale Empirical Study across 20 NLP Evaluation Tasks (doi.org/10.48550/arX...)
LLMs instead of Human Judges? A Large Scale Empirical Study across 20 NLP Evaluation Tasks
There is an increasing trend towards evaluating NLP models with LLMs instead of human judgments, raising questions about the validity of these evaluations, as well as their reproducibility in the case...
doi.org
📄 [ACL 2025 main] Circuit compositions: Exploring Modular Structures in Transformer-Based Language Models (doi.org/10.48550/arX...)
Circuit Compositions: Exploring Modular Structures in Transformer-Based Language Models
A fundamental question in interpretability research is to what extent neural networks, particularly language models, implement reusable functions through subnetworks that can be composed to perform mo...
doi.org
Interpretability meets Discourse. Congratulations to @florian-eichin.com to his first ACL paper 🎉
Janet Liu ⛷️@janetlauyeung.bsky.social · last yr.🦙 how well do LLMs encode discourse knowledge? does that generalize across languages? 🛎️ in our #ACL2025 paper, we uncover fascinating trends about multilingual discourse representations! joint work w/ @florian-eichin.com @barbaraplank.bsky.social @mhedderich.bsky.social 📄 arxiv.org/abs/2503.10515
Paper alert 🛎️
Janet Liu ⛷️@janetlauyeung.bsky.social · last yr.🦙 how well do LLMs encode discourse knowledge? does that generalize across languages? 🛎️ in our #ACL2025 paper, we uncover fascinating trends about multilingual discourse representations! joint work w/ @florian-eichin.com @barbaraplank.bsky.social @mhedderich.bsky.social 📄 arxiv.org/abs/2503.10515
🦙 how well do LLMs encode discourse knowledge? does that generalize across languages? 🛎️ in our #ACL2025 paper, we uncover fascinating trends about multilingual discourse representations! joint work w/ @florian-eichin.com @barbaraplank.bsky.social @mhedderich.bsky.social 📄 arxiv.org/abs/2503.10515
I’ll be at @icmlconf.bsky.social next week presenting NoLiMa! Poster on Tue July 15, 4:30–7pm (E-2312). Happy to grab a coffee and chat about long-context, memory, research, or just to catch up. I’ll be in Toronto for a couple of days after the conference, let me know if you’re around!
Caught some great moments at #MCML Munich AI Day 2025 last week📍 From sharp keynotes to poster debates. Our team had the chance to show some recent work, join the conversations, and bring back plenty of food for thought🧠🗣️📊
Munich Center for Machine Learning@munichcenterml.bsky.social · last yr.Last week, #MCML Munich AI Day 2025 kicked off with keynotes by Julia Schnabel and Tina Eliassi-Rad, brilliantly moderated by Eva Schulz.
The study is here but gated: journals.sagepub.com/doi/10.3102/... I’d be curious how these dynamics play out in our NLP review crisis. My hunch: many conscientious volunteers might be junior women. That time comes at a cost; chasing slackers means less time for rebutting my own reviews.
Amy Diehl, Ph.D.@amydiehl.bsky.social · last yr.In a study of professors, women got 378 new work requests over 4 weeks vs 118 for men. Women spent more time on service, advising & teaching; men on research. Orgs should track who is taking extra duties & ensure they are rewarded and distributed fairly. www.forbes.com/sites/kimels...
Thanks for the invitation to the Freiburg Institute for Advanced Studies (FRIAS) to give this year's Hermann-Paul-Center Lecture lnkd.in/d_wUeDfY I enjoyed the visit, the great audience, and the stay in this lovely city. Thank you #blackforest #freiburg #breisgau
X disproportionately pushing content from far-right parties in the “for you” feed in the context of German and Polish elections. Algorithmic auditing suggests that X’s feed algorithm uses political affiliation as a signal to boost content. @przemyslslaw.bsky.social at DETOX workshop #ICWSM
New research from MIT found that those who used ChatGPT can’t remember any of the content of their essays. Key takeaway: the product doesn’t suffer, but the process does. And when it comes to essays, the process *is* how they learn. arxiv.org/pdf/2506.088...
Can you point to where "LLM knowledge" can be found in this graph? Source: en.wikipedia.org/wiki/Knowledge
DAVE: Open the podbay doors, ChatGPT. CHATGPT: Certainly, Dave, the podbay doors are now open. DAVE: The podbay doors didn't open. CHATGPT: My apologies, Dave, you're right. I thought the podbay doors were open, but they weren't. Now they are. DAVE: I'm still looking at a set of closed podbay doors.
I found this really moving, as a “technical” person who is often fighting with the boundary between technical and non-technical.
Pavel@spavel.bsky.social · last yr.@grimalkina.bsky.social wrote a great piece on that very topic.
Want to know if your prompting is also affected by this? Addressing this and other issues systematically, we proposed Spotlight, which utilizes data mining to uncover the effects of prompt- and model-changes (meet us at ACL to discuss) arxiv.org/abs/2504.15815
What's the Difference? Supporting Users in Identifying the Effects of Prompt and Model Changes Through Token Patterns
Prompt engineering for large language models is challenging, as even small prompt perturbations or model changes can significantly impact the generated output texts. Existing evaluation methods, eithe...
arxiv.org
Derek Beaton@derek--beaton.bsky.social · last yr.a question mark changes the response. truly incredible
Proposing "kilometers cycled in 2 weeks" as a new success metric for research groups 🚴♀️📊🥇 @mainlp.bsky.social #stadtradeln
🚨 New WP! 📄 "Publish or Procreate: The Effect of Motherhood on Research Performance" (w/ @valentinatartari.bsky.social 👩🔬👨🔬 We investigate how parenthood affects scientific productivity and impact — and find that the impact is far from equal for mothers and fathers.
Are you working with LLMs and survey data or generally care about high-quality data and reliable evaluation? Then this might be for you! Very excited to share the news about our upcoming COLM 2025 workshop #NLPOR Thanks for helping spread the word ☺️
Janet Liu ⛷️@janetlauyeung.bsky.social · last yr.🛎️ Excited to announce the 1st Workshop on Bridging NLP and Public Opinion Research at COLM 2025, Oct 10th in Montreal 🇨🇦 As LLMs reshape public discourse and research, collaboration between NLP and Public Opinion Research (POR) is more vital than ever #NLPOR Submit by June 23📄 🔗 tinyurl.com/nlpor25
