←
Magda Dubois95 posts
Magda Dubois avatar

Magda Dubois

@DubMagda

Research Scientist @AISecurityInst working on LLM evaluations 🤖 | PhD in computational cognitive neuroscience @MPC_CompPsych 🧠

◷ Joined December 2016▧ 15 Photos and videos
Cozmin Ududec avatar
⟳ Magda Dubois retweeted
Cozmin Ududec@CUdudec·Feb 26

New from the Science of Evaluation Team at @AISafetyInst: a pipeline for rigorous transcript analysis. I think transcript analysis is still underrated, especially as model horizons are getting longer and task environments more complex.

Arvindh Arun avatar
⟳ Magda Dubois retweeted
Arvindh Arun@arvindh__a·12 Sep 2025

Why does horizon length grow exponentially as shown in the METR plot? Our new paper investigates this by isolating the execution capabilities of LLMs. Here's why you shouldn't be fooled by slowing progress on typical short-task benchmarks... 🧵

Post attachment
Konrad Rieck 🌈 avatar
⟳ Magda Dubois retweeted
Konrad Rieck 🌈@mlsec·1 Jul 2025

We're excited to announce the Call for Papers for SaTML 2026, the premier conference on secure and trustworthy machine learning @satml_conf We seek papers on secure, private, and fair learning algorithms and systems. 👉 satml.org/call-for-papers/ ⏰ Deadline: Sept 24

Post attachment
Sahar Abdelnabi 🕊 avatar
⟳ Magda Dubois retweeted
Sahar Abdelnabi 🕊@sahar_abdelnabi·1 Jun 2025

Hawthorne effect describes how study participants modify their behavior if they know they are being observed In our paper 📢, we study if LLMs exhibit analogous patterns🧠 Spoiler: they do⚠️ 🧵1/n

Post attachment
summerfieldlab @summerfieldlab.bsky.social avatar
⟳ Magda Dubois retweeted
summerfieldlab @summerfieldlab.bsky.social@summerfieldlab·9 Jul 2025

In a new paper, we examine recent claims that AI systems have been observed ‘scheming’, or making strategic attempts to mislead humans. We argue that to test these claims properly, more rigorous methods are needed.

Post attachment
AI Security Institute (AISI) avatar
⟳ Magda Dubois retweeted
AI Security Institute (AISI)@AISecurityInst·9 Jul 2025

Evaluating AI models is essential for improving their performance and understanding their risks. Increasingly, researchers are using “autograders” – having Large Language Models (LLMs) grade model outputs. But how do we know if these autograders are reliable? 🧵

Magda Dubois avatar
Magda Dubois@DubMagda·13 May 2025

New paper introducing a framework to better quantify uncertainty in LLM evaluations (led by @LLuettgau🙌). A beta Python package (developed by @HarryCoppock🚀) is available if you want to try it out. ➡️Get in touch if you have any Qs/feedback! Paper: arxiv.org/abs/2505.05602

Advanced AI systems require complex evaluations to measure abilities, but conventional analysis techniques often fall short. Introducing HiBayES: a flexible, robust statistical modelling framework that accounts for the nuances & hierarchical structure of advanced evaluations.

Post attachment
Lennart Luettgau avatar
⟳ Magda Dubois retweeted
Lennart Luettgau@LLuettgau·23 Sep 2024

Excited to share our brand-new work shedding some light on the neural mechanisms behind one of human’s coolest cognitive feats: compositional generalization of structural knowledge! A Tweeprint-Thread 🧵 1/n

Alexandr Wang avatar
⟳ Magda Dubois retweeted
Alexandr Wang@alexandr_wang·25 Jul 2024

1/ New paper in Nature shows model collapse as successive model generations models are recursively trained on synthetic data. This is an important result. While many researchers today view synthetic data as AI philosopher’s stone, there is no free lunch. Read more 👇

Post attachment
Matthew Nour avatar
⟳ Magda Dubois retweeted
Matthew Nour@Matt_Nour·10 Oct 2023

Paper out in @PNASNews! A 'cognitive mapping' lens on language in psychosis, using word embedding models, computational modelling, and MEG. A hint of what's to come at @OxPsychiatry and @UCLBrainScience... With @mcneural_, @YunzheNeuro, Ray Dolan. pnas.org/doi/abs/10.1073/pna…

Lennart Luettgau avatar
⟳ Magda Dubois retweeted
Lennart Luettgau@LLuettgau·1 Sep 2023

Preprint alert🚨! In this new paper we study how humans decompose dynamical subprocesses and leverage the abstracted subprocesses for compositional reuse of experience in new situations. psyarxiv.com/sxn4a/ Tweeprint to follow soon!

Marcelo Mattar avatar
⟳ Magda Dubois retweeted
Marcelo Mattar@marcelomattar·4 May 2023

In our lab's latest paper, we introduce a novel modeling approach using RNNs to reveal the cognitive algorithms behind animal decision-making. Check out our preprint, led by UCSD PhD student @Ji_An_Li and co-authored by Marcus Benna: biorxiv.org/content/10.1101/…

Magda Dubois avatar
Magda Dubois@DubMagda·6 Apr 2023

Congratulations to my academic sibling @AlisaLoosen for those (very) well-deserved three shiny balloons

Post attachment
Magda Dubois avatar
Magda Dubois@DubMagda·14 Mar 2023

Wanna try out a (cool🦙) alternative to GPT?

Yann Dubois@yanndubs13 Mar 2023

🦙Excited to share this demo of Alpaca 🔥Highlights: ~GPT3.5 performance for < 600$🔥 The goal was to have a simple model /training procedure that academics could study and improve with limited resources We achieved that by finetuning a 7B LLaMA on 52K generated instructions