Profile avatar
ayukh.bsky.social
Statistics MSc @ ETH Zurich Multilingual LLM training/eval/safety MSc soldier @ SRI lab ayukh.com
17 posts 121 followers 469 following
Regular Contributor
Posts 15 Comments 7

EuroLLM HF space test prompts are very interesting 🤔

submitted 6 days ago • 0 comments

1/1 complete

submitted 77 days ago • 0 comments

📢 Our benchmark on self-supervised learning for single-cell data🧬 is accepted at the #NeurIPS2024 SSL workshop. We take a first step towards establishing best practices for SSL methods for single-cell data, and benchmark 8 SSL methods on 3 downstream tasks across 8 datasets.

submitted 78 days ago • 1 comment

Watch me stalk Kaggle in Vancouver to get stickers

submitted 83 days ago • 1 comment

Is MMLU Western-centric? 🤔 As part of a massive cross-institutional collaboration: 🗽Find MMLU is heavily overfit to western culture 🔍 Professional annotation of cultural sensitivity data 🌍 Release improved Global-MMLU 42 languages 📜 Paper: arxiv.org/pdf/2412.03304 📂 Data: hf.co/datasets/Coh...

submitted 85 days ago • 7 comments

Going to #neurips2024 next week

submitted 85 days ago • 0 comments

Цей день настав - Ukrainian is finally recognized as a mid-resource language 🇺🇦🦅🦅🦅

submitted 86 days ago • 1 comment

Exciting end of the year! - Won the GraySwanAI jailbreaking challenge for harmful code generation - Proud Ukrainian ambassador for Cohere4AI new Aya Expanse models - Started my master thesis👩‍🍳🇺🇦👀 - Going to Vancouver🇨🇦 for @neuripsconf.bsky.social to chat about LLM privacy and SynthPAI #neurips

submitted 91 days ago • 0 comments

Seems to be a hot take here: why are people getting mad about stuff they post *themselves* online? Your posts online are getting scraped all the time and you chose an open-source movement leader as a scapegoat. Just initiate a discussion with HF about the "right to be forgotten" from GDPR

submitted 92 days ago • 0 comments

- get compute to GPT go brr - get banned twice by GCP (they thought I was mining crypto) - models do not fit on local disks - 4300 training hours for 1 epoch - I need to learn JAX after all

submitted 95 days ago • 0 comments

Zaporizhzhia, a big Ukrainian city of almost one million, is under a massive drone attack. terrorism is russian culture.

submitted 101 days ago • 11 comments

One year of ChatGPT has shown incredible capabilities of LLMs. However, they still have lots of problems! The LVE project aims at addressing this - with LVEs we track LLM vulnerabilities and exposures in an open-source community-first approach. Contribute and more info: lve-project.org #NLP #LLM

submitted 443 days ago • 0 comments

submitted 484 days ago • 1 comment

Тред з корисними порадами для тих, хто щойно долучився. 🧵

submitted 643 days ago • 62 comments

Купила українські вареники в Швейцарії

submitted 599 days ago • 1 comment