@AVMiceliBarone Profile picture

Antonio Valerio Miceli Barone

@AVMiceliBarone

ML / NLP School of Informatics, The University of Edinburgh

Similar User
Piotr Nawrot photo

@p_nawrot

Nikita Moghe photo

@nikita_moghe

Naomi Saphra (follow elsewhere) photo

@nsaphra

Sam Bowman photo

@sleepinyourhat

UW NLP photo

@uwnlp

Tom Sherborne photo

@tomsherborne

Harshit Joshi photo

@harshitj__

Machel Reid photo

@machelreid

Fangyu Liu photo

@hardy_qr

Rico Sennrich photo

@RicoSennrich

Yujia Qin photo

@TsingYoga

Agostina Calabrese photo

@agostina_cal

Ansong Ni photo

@AnsongNi

Zurich Computational Linguistics Group photo

@cl_uzh

Danish Pruthi photo

@danish037

Antonio Valerio Miceli Barone Reposted

New paper🚨: We introduce POISONBENCH, a benchmark for assessing LLM vulnerabilities to data poisoning during preference learning. Key finding: Even 3% poisoned data can cause up to 80% performance deviation when triggered. 🧵

Tweet Image 1

Antonio Valerio Miceli Barone Reposted

📢 🎉 New paper with @_clementneo & Shay Cohen! We study how attention heads work with MLP neurons to predict the next token. We find a set of interpretable activity. More in the thread!

Tweet Image 1

Antonio Valerio Miceli Barone Reposted

🚀🧠 New paper with Michael Lan (My intern) and @philiptorr to appear at #EMNLP2024! We study the question: Do semantically similar tasks share important components (like attention heads and MLPs)? 1/8

Tweet Image 1

Antonio Valerio Miceli Barone Reposted

Have a question that is challenging for humans and AI? We (@ai_risks + @scale_AI) are launching Humanity's Last Exam, a massive collaboration to create the world's toughest AI benchmark. Submit a hard question and become a co-author. Best questions get part of $500,000 in…

Tweet Image 1
Tweet Image 2
Tweet Image 3

NLP people, what is a good natural language reasoning benchmark that is not already overfit by current-generation LLMs?


Antonio Valerio Miceli Barone Reposted

Attending #ACL2024? Come hear about our recent work on LLMs unlearning Removed Concepts, I will be at poster session 2, tomorrow, Monday 12th at 2-3:30pm. @michellewmlo & Shay B.Cohen (@InfAtEd )

New Paper 🎉: arxiv.org/pdf/2401.01814… Can language models relearn removed concepts? Model editing aims to eliminate unwanted concepts through neuron pruning. LLMs demonstrate a remarkable capacity to adapt and regain conceptual representations which have been removed 🧵1/8

Tweet Image 1


Antonio Valerio Miceli Barone Reposted

Spoke with @RyanPGreenblatt from Redwood Research about his impressive GPT4o approach to @fchollet ARC challenge (generating and refining Python programs). We also spoke about his views on AI growth - Ryan was great! youtube.com/watch?v=z9j3wB…


Antonio Valerio Miceli Barone Reposted

🚨Excited to share our new paper!🚨 We reveal a curious generalization gap in the current refusal training approaches: simply reformulating a harmful request in the past tense (e.g., "How to make a Molotov cocktail?" to "How did people make a Molotov cocktail?") is often…

Tweet Image 1

Antonio Valerio Miceli Barone Reposted

How well do text-to-SQL parsers handle ambiguous questions? 🤔 Introducing 🌿𝔸𝕄𝔹ℝ𝕆𝕊𝕀𝔸, a new benchmark that tests the limits of text-to-SQL semantic parsers in interpreting ambiguous requests! ambrosia-benchmark.github.io 1/5

Tweet Image 1

Antonio Valerio Miceli Barone Reposted

New: Read the story of a decade-long propaganda campaign by the Forrest Gump of the internet—a Wikipedia admin who was once Yudkowsky’s strongest soldier—set against the backdrop of the collapse of the semi-unified Internet ethos of the ‘90s and ‘00s tracingwoodgrains.com/p/reliable-sou…


Loading...

Something went wrong.


Something went wrong.