Matthew Muckley
@mattmucklm
Followers
970
Following
775
Statuses
226
Research Engineer, Meta Fundamental AI Research (FAIR). ML for compression, computer vision, medicine. Threads: https://t.co/IwcbQ8VDPn
New York, NY
Joined December 2018
RT @brandondamos: If you prompt an LLM and stop in the middle of a token, what happens? ❌ The generated response doesn't correctly complete…
0
7
0
Tokenization is a limitation of modern LLMs. How to make it better? Find a way to convert your LLM's probabilities to byte-level probabilities! Details in the thread below! ⬇️ Work done by our great intern @buutphan
🤔Tokenization (1) makes your LLMs produce odd text when prompts are cut off mid-token? (2) gives you problems when ensembling different LLMs? 💡We solve both by converting tokenized LLMs into equivalent byte-level LLMs! No training required! 📎Paper:
1
0
11
One example here by the responsible AI team!
Our responsible AI team is hiring 3 research scientist interns this cycle (2 in Montreal, one in NYC). We're seeking enrolled PhD students who are excited to spend their summer figuring out how to ensure vision and/or language models work for everyone!
0
0
0
@avinab_saha There are a lot more legal restrictions around generative diffusion models, and it took too for us long to resolve before the first author finished their contract. However, an independent group has released an open implementation with some improvements:
1
0
0
@rohitrango There are some projects that will touch on health/medicine (e.g., the DINO team with , but to my knowledge there are no teams where that is their core research.
1
0
1
0
0
0
RT @karen_ullrich: Even with preference alignment, LLMs can be enticed into harmful behavior via adversarial prompts 😈. 🚨 Breaking: our t…
0
20
0
Appearing at ICML this week!
Preprint and code of my internship @Meta on neural-augmented residual quantization is now online: ⚡️ 🚨We heavily improve SOTA compression and search performance by conditioning the codebook in each residual quantization step on selected codewords so far.
0
0
4
RT @buutphan: Why do LLMs fail simple completion tasks, but not on a harder task? Learn about tokenization bias in LLMs and how to fix it…
0
11
0
RT @Piovrasca: Are sota image generative models effective world models? Consistency-diversity-realism Pareto fronts show they're not (yet)…
0
42
0
@TacoCohen Thank you for articulating why machine learning papers are so confusing for us image processing people.
0
0
3