Marius Mosbach
mariusmosbach.bsky.social
Marius Mosbach
@mariusmosbach.bsky.social
#NLP Postdoc at Mila - Quebec AI Institute & McGill University

mariusmosbach.com
Reposted by Marius Mosbach
Our new paper in #PNAS (bit.ly/4fcWfma) presents a surprising finding—when words change meaning, older speakers rapidly adopt the new usage; inter-generational differences are often minor.

w/ Michelle Yang, ‪@sivareddyg.bsky.social‬ , @msonderegger.bsky.social‬ and @dallascard.bsky.social‬👇(1/12)
July 29, 2025 at 12:06 PM
Reposted by Marius Mosbach
🚨Job Alert
W2 (TT W3) Professorship in Computer Science "AI for People & Society"
@saarland-informatics-campus.de/@uni-saarland.de is looking to appoint an outstanding individual in the field of AI for people and society who has made significant contributions in one or more of the following areas:
July 18, 2025 at 7:11 AM
Reposted by Marius Mosbach
📣 Life update: Thrilled to announce that I’ll be starting as faculty at the Max Planck Institute for Software Systems this Fall!

I’ll be recruiting PhD students in the upcoming cycle, as well as research interns throughout the year: lasharavichander.github.io/contact.html
July 22, 2025 at 4:12 AM
Reposted by Marius Mosbach
I'm at #ICML in Vancouver this week, hit me up if you want to chat about pre-training experiments or explainable machine learning.

You can find me at these posters:

Tuesday: How Much Can We Forget about Data Contamination? icml.cc/virtual/2025...
July 14, 2025 at 2:49 PM
Reposted by Marius Mosbach
Mechanistic interpretability often relies on *interventions* to study how DNNs work. Are these interventions enough to guarantee the features we find are not spurious? No!⚠️ In our new paper, we show many mech int methods implicitly rely on the linear representation hypothesis🧵
July 14, 2025 at 12:15 PM
Reposted by Marius Mosbach
Have you ever wondered whether a few times of data contamination really lead to benchmark overfitting?🤔 Then our latest #ICML paper about the effect of data contamination on LLM evals might be for you!🚀

Paper: arxiv.org/abs/2410.03249
👇🧵
July 8, 2025 at 6:42 AM
Reposted by Marius Mosbach
💡Beyond math/code, instruction following with verifiable constraints is suitable to be learned with RLVR.
But the set of constraints and verifier functions is limited and most models overfit on IFEval.
We introduce IFBench to measure model generalization to unseen constraints.
July 3, 2025 at 9:06 PM
Reposted by Marius Mosbach
A blizzard is raging through Montreal when your friend says “Looks like Florida out there!” Humans easily interpret irony, while LLMs struggle with it. We propose a 𝘳𝘩𝘦𝘵𝘰𝘳𝘪𝘤𝘢𝘭-𝘴𝘵𝘳𝘢𝘵𝘦𝘨𝘺-𝘢𝘸𝘢𝘳𝘦 probabilistic framework as a solution.
Paper: arxiv.org/abs/2506.09301 to appear @ #ACL2025 (Main)
June 26, 2025 at 3:52 PM
Reposted by Marius Mosbach
Started a new podcast with @tomvergara.bsky.social !

Behind the Research of AI:
We look behind the scenes, beyond the polished papers 🧐🧪

If this sounds fun, check out our first "official" episode with the awesome Gauthier Gidel
from @mila-quebec.bsky.social :

open.spotify.com/episode/7oTc...
02 | Gauthier Gidel: Bridging Theory and Deep Learning, Vibes at Mila, and the Effects of AI on Art
Behind the Research of AI · Episode
open.spotify.com
June 25, 2025 at 3:54 PM
Reposted by Marius Mosbach
Interested in shaping the progress of responsible AI and meeting leading researchers in the field? SoLaR@COLM 2025 is looking for paper submissions and reviewers!

🤖 ML track: algorithms, math, computation
📚 Socio-technical track: policy, ethics, human participant research
June 17, 2025 at 5:46 PM
Reposted by Marius Mosbach
"Build the web for agents, not agents for the web"

This position paper argues that rather than forcing web agents to adapt to UIs designed for humans, we should develop a new interface optimized for web agents, which we call Agentic Web Interface (AWI).

arxiv.org/abs/2506.10953
June 14, 2025 at 4:17 AM
Reposted by Marius Mosbach
Excited to share the results of my recent internship!

We ask 🤔
What subtle shortcuts are VideoLLMs taking on spatio-temporal questions?

And how can we instead curate shortcut-robust examples at a large-scale?

We release: MVPBench

Details 👇🔬
June 13, 2025 at 2:47 PM
Reposted by Marius Mosbach
New paper in Interspeech 2025 🚨
@interspeech.bsky.social

A Robust Model for Arabic Dialect Identification using Voice Conversion

Paper 📝 arxiv.org/pdf/2505.24713
Demo 🎙️https://shorturl.at/rrMm6

#Arabic #SpeechTech #NLProc #AI #Speech #ArabicDialects #Interspeech2025 #ArabicNLP
June 10, 2025 at 10:07 AM
Reposted by Marius Mosbach
Do LLMs hallucinate randomly? Not quite.

Our #ACL2025 (Main) paper shows that hallucinations under irrelevant contexts follow a systematic failure mode — revealing how LLMs generalize using abstract classes + context cues, albeit unreliably.

📎 Paper: arxiv.org/abs/2505.22630 1/n
June 6, 2025 at 6:10 PM
Reposted by Marius Mosbach
Chain-of-Thought (CoT) reasoning lets LLMs solve complex tasks, but long CoTs are expensive. How short can they be while still working? Our new ICML paper tackles this foundational question.
May 5, 2025 at 12:25 PM
Reposted by Marius Mosbach
Come to my keynote tomorrow at the first official @queerinai.com workshop at #NAACL2025 to hear about how trans languaging is complex and cool, and how this makes it extra difficult to process computationally. I will have SO many juicy examples!
May 3, 2025 at 8:52 PM
Reposted by Marius Mosbach
Deadline extended! ⏳

The Actionable Interpretability Workshop at #ICML2025 has moved its submission deadline to May 19th. More time to submit your work 🔍🧠✨ Don’t miss out!
May 3, 2025 at 8:00 PM
Check out Gaurav's video on their #NAACL paper and find @adadtur.bsky.social at the conference 👇
Congratulations to Mila members @adadtur.bsky.social , Gaurav Kamath and @sivareddyg.bsky.social for their SAC award at NAACL! Check out Ada's talk in Session I: Oral/Poster 6. Paper: arxiv.org/abs/2502.05670
May 2, 2025 at 1:41 AM
Reposted by Marius Mosbach
I'll be at #NAACL2025:

🖇️To present my paper "Superlatives in Context", showing how the interpretation of superlatives is very context dependent and often implicit, and how LLMs handle such semantic underspecification

🖇️And we will present RewardBench on Friday

Reach out if you want to chat!
April 27, 2025 at 8:00 PM
Reposted by Marius Mosbach
I’m really excited about Diffusion Steering Lens, an intuitive and elegant new “logit lens” technique for decoding the attention and MLP blocks of vision transformers!

Vision is much more expressive than language, so some new mech interp rules apply:
🔍Logit Lens tracks what transformer LMs “believe” at each layer. How can we effectively adapt this approach to Vision Transformers?

Happy to share our “Decoding Vision Transformers: the Diffusion Steering Lens” was accepted at the CVPR 2025 Workshop on Mechanistic Interpretability for Vision!
(1/7)
April 25, 2025 at 1:36 PM
Reposted by Marius Mosbach
💡 New ICLR paper! 💡
"On Linear Representations and Pretraining Data Frequency in Language Models":

We provide an explanation for when & why linear representations form in large (or small) language models.

Led by @jackmerullo.bsky.social, w/ @nlpnoah.bsky.social & @sarah-nlp.bsky.social
April 25, 2025 at 1:55 AM
Reposted by Marius Mosbach
Lots of progress in mech interp (MI) lately! But how can we measure when new mech interp methods yield real improvements over prior work?

We propose 😎 𝗠𝗜𝗕: a 𝗠echanistic 𝗜nterpretability 𝗕enchmark!
April 23, 2025 at 6:15 PM
Paper title of the year so far. I will be back ... have to read the paper now. Great work @saxon.me !
Check out our new paper on benchmarking and mitigating overthinking in reasoning models!

From a simple observational measure of overthinking, we introduce Thought Terminator, a black-box, training-free decoding technique where RMs set their own deadlines and follow them

arxiv.org/abs/2504.13367
April 21, 2025 at 11:43 PM
Checkout Benno's notes about our impact of interpretability paper 👇.

Also, we are organizing a workshop at #ICML2025 which is inspired by some of the questions discussed in the paper: actionable-interpretability.github.io
April 15, 2025 at 11:11 PM
Our Thoughtology 💭 paper finally made it to arXiv (after being on hold for more than a week 😵‍💫). Make sure to check it out if you are interested in analyzing reasoning chains of LLMs.
🔗: arxiv.org/abs/2504.07128
April 11, 2025 at 12:45 PM