Yanai Elazar @ACL

A strange trend I've noticed at #ACL2025 is that people are hesitant to reach out to papers/"academic products" authors.
This is unfortunate for both parties! A simple email can save a lot of time to the sender, but is also one of my favorite kind of email as the receiver!

EPFL NLP Postdoctoral Scholar Posting - Swiss AI LLMs The EPFL Natural Language Processing (NLP) lab is looking to hire a postdoctoral researcher candidate in the area of multilingual LLM design, training, and evaluation. This postdoctoral position is as...

The EPFL NLP lab is looking to hire a postdoctoral researcher on the topic of designing, training, and evaluating multilingual LLMs:

docs.google.com/document/d/1...

Come join our dynamic group in beautiful Lausanne!

Had a really great and fun time with @yanai.bsky.social, Niloofar Mireshghallah, and Reza Shokri discussing memorisation at the @l2m2workshop.bsky.social panel. Thanks to the entire organising team and attendees for making this such a fantastic workshop! #ACL2025

I just wish a particular conversation from the workshop's dinner would be appended to the panel's proceedings

I had a lot of fun contemplating about memorization questions at the @l2m2workshop.bsky.social panel yesterday together with Niloofar Mireshghallah and Reza Shokri, moderated by
@pietrolesci.bsky.social who did a fantastic job!
#ACL2025

I'll present our work w/ @santosh-tyss.bsky.social
@yanai.bsky.social @barbaraplank.bsky.social on LLMs memorization of distributions of political leanings in their pretraining data! Catch us at L2M2 workshop @l2m2workshop.bsky.social #ACL2025 tmrw
📆 Aug 1, 14:00–15:30 📑 arxiv.org/pdf/2502.18282

Ai2 is excited to be at #ACL2025 in Vienna, Austria this week. Come say hello, meet the team, and chat about the future of NLP. See you there! 🤝📚

It's crazy that people give more than a single invited talk during the same conference (diff workshops).
A single talk (done right) is challenging enough.

😂

What if you could understand and control an LLM by studying its *smaller* sibling?

Our new paper introduces the Linear Representation Transferability Hypothesis. We find that the internal representations of different-sized models can be translated into one another using a simple linear(affine) map.

I really like this take. Academia and the open source community should embrace transparency in data, even in the cost of the issues that come with it. These issues should be of course studied and documented, but not used as an indicator or a signal to shut down the whole operation.

Prompting is our most successful tool for exploring LLMs, but the term evokes eye-rolls and grimaces from scientists. Why? Because prompting as scientific inquiry has become conflated with prompt engineering.

This is holding us back. 🧵and new paper with @ari-holtzman.bsky.social .

First Align, then Predict: Understanding the Cross-Lingual Ability of Multilingual BERT Multilingual pretrained language models have demonstrated remarkable zero-shot cross-lingual transfer capabilities. Such transfer emerges by fine-tuning on a task of interest in one language and evalu...

Very cool!
Reminds me of these "old" papers where we found language agnostic representations in early layers of multilingual models:
arxiv.org/abs/2101.11109
arxiv.org/abs/2010.08275

Chain-of-Thought Is Not Explainability | alphaXiv View recent discussion. Abstract: Chains-of-thought (CoT) allow language models to verbalise multi-step rationales before producing their final answer. While this technique often boosts task performan...

arxiv refused to upload this paper for some reason..
we have it on alphaxiv for now - www.alphaxiv.org/abs/2025.02

What's up with @arxiv-cs-cl.bsky.social
Wasn't the entire premise of this website to allow uploading of papers w/o the official peer review process??

Check out our take on Chain-of-Thought.
I really like this paper as a survey on the current literature on what CoT is, but more importantly on what it's not.
It also serves as a cautionary tale to the (apparently quite common) misuse of CoT as an interpretable method.

🥲

But that's true for all conferences right? I don't think anything meaningful changed with that regard since ARR

@annarogers.bsky.social (et al) did a good job at that at ACL 23. I think the problem is not the lack of instructions, but that people tend to not read them..

מלקולם גלדוול נהדר

Congrats!!

"the highest bar for scientific work in the field"

Applying Intrinsic Debiasing on Downstream Tasks: Challenges and Considerations for Machine Translation Most works on gender bias focus on intrinsic bias -- removing traces of information about a protected group from the model's internal representation. However, these works are often disconnected from t...

We show that these techniques are quite limited, in terms of their downstream effects (here on translation):
arxiv.org/abs/2406.00787

𝐇𝐨𝐰 𝐜𝐚𝐧 𝐰𝐞 𝐩𝐞𝐫𝐟𝐞𝐜𝐭𝐥𝐲 𝐞𝐫𝐚𝐬𝐞 𝐜𝐨𝐧𝐜𝐞𝐩𝐭𝐬 𝐟𝐫𝐨𝐦 𝐋𝐋𝐌𝐬?

Our method, Perfect Erasure Functions (PEF), erases concepts perfectly from LLM representations. We analytically derive PEF w/o parameter estimation. PEFs achieve pareto optimal erasure-utility tradeoff backed w/ theoretical guarantees. #AISTATS2025 🧵

Cool! Will check it out.

Congrats!! Great paper! 🥳👏

Congrats!! Great paper 🥳👏

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!

I'm in Singapore for ICLR to present this paper:
Tomorrow, April 26th, 10-12:30 in Hall 3+2B #236
Come check it out!

arxiv.org/abs/2504.12459

I think this is super exciting as it shows a way to use interpretability metrics as a useful feature to explain, and reveal information about the training data, even without having access to it (see some examples in the attached table).