Matthew Finlayson
@mattf1n
PhD at @nlp_usc | Former predoc at @allen_ai on @ai2_aristo | Harvard 2021 CS & Linguistics
Wanna know gpt-3.5-turbo's embed size? We find a way to extract info from LLM APIs and estimate gpt-3.5-turbo’s embed size to be 4096. With the same trick we also develop 25x faster logprob extraction, audits for LLM APIs, and more! 📄 arxiv.org/abs/2403.09539 Here’s how 1/🧵
this was an amazing project. Matt is an absolute joy to work with with his ever extending support and his genius ideas. and yes, hire me :). my dms are open.
The project was led by @murtazanazir, an independent researcher with serious engineering chops. It's his first paper. He's a joy to work with and is applying to PhDs. Hire him! It's great to finally collab with @jxmnop, and a big thanks to @swabhz and @xiangrenNLP for advising.
excited to finally share this paper. still shocked that this works so well! this was a fun project with matt, @jxmnop, @swabhz and @xiangrenNLP . checkout the demo on matt's blog. feel free to reach out for any issues!
I didn't believe when I first saw, but: We trained a prompt stealing model that gets >3x SoTA accuracy. The secret is representing LLM outputs *correctly* 🚲 Demo/blog: mattf1n.github.io/pils 📄: arxiv.org/abs/2506.17090 🤖: huggingface.co/dill-lab/pils-… 🧑💻: github.com/dill-lab/PILS
Loving the #NeurIPS2024 'Beyond Decoding: Meta-Generation Algorithms for LLMs' workshop ❤️ by @wellecks @mattf1n @haileysch__: 1. Primitive generators: optimization vs sampling 2. Meta-generators: chain, parallel, tree, refinement 3. Efficiency: quantize, FA, sparse MoEs, KV…
I didn’t realize when making these diagrams that my Taylor example would be so timely 😂
In Vancouver for NeurIPS but don't have Taylor Swift tickets? You can still spend the day going through our tutorial reading list: - cmu-l3.github.io/neurips2024-in… Tuesday December 10, 1:30-4:00pm @ West Exhibition Hall C, NeurIPS
We're incredibly honored to have an amazing group of panelists: @agarwl_ , @polynoamial , @BeidiChen, @nouhadziri, @j_foerst , with @uralik1 moderating We'll close with a panel discussion about scaling, inference-time strategies, the future of LLMs, and more!
Curious about inference-time scaling, the #1 trending topic in LLMs? Come to our NeurIPS tutorial: Beyond Decoding: Meta-Generation Algorithms for LLMs (Tue. @ 1:30)! cmu-l3.github.io/neurips2024-in…
Excited to give a NeurIPS tutorial on LLM inference strategies, inference-time scaling laws & more with @mattf1n and @haileysch__ ! "Beyond Decoding: Meta-Generation Algorithms for Large Language Models" More details soon, check out arxiv.org/abs/2406.16838 in the meantime!
We’re excited to share the list of accepted tutorials for @NeurIPSConf ! Thanks to everyone who put in the time to submit a proposal. Check out the lineup and let us know which tutorials you’re most looking forward to! blog.neurips.cc/2024/10/17/int… with @irenetrampoline & @GalChechik
Thank you so much @SpecNews1SoCal @jaskang21 for featuring our work on OATH-Frames: Characterizing Online Attitudes towards Homelessness with LLM Assistants👇 🖥️📈 oath-frames-dashboard.streamlit.app 🗞️ spectrumnews1.com/ca/southern-ca… @CSatUSC @nlp_usc @uscsocialwork @CAIS_USC @USCViterbi @swabhz
Arrived in Philadelphia for the very 1st @COLM_conf! Excited to catch up w/ everyone & happy to chat about faculty/phd positions @USCViterbi 🙂 Plz meet our amazing PhD students (@huihan_li @mattf1n @sahana_ramnath ) for their work on model safety and cultural bias analysis 👇
I had a fantastic time visiting USC and talking about 🌎AppWorld (appworld.dev) last Friday!! Thank you, @swabhz, @mattf1n, & @BrihiJ, for inviting and hosting me. Also thank you, @robinomial, @_jessethomason_, & many others, for insightful discussions and meetings!
📢 I am giving a talk on 🌎 AppWorld & its future works at 13+ universities (USC, UCI, Stanford, Berkeley, Princeton, JHU, ..) and companies (Ai2, Google, Apple, Semantic Machines, ..) in the next 1-2 months 📅 Schedule+Details: appworld.dev/talks
Just landed in Philly for @COLM_conf where I’ll be presenting my work on extracting secrets from LLM APIs at the Wednesday afternoon poster sesh. Please reach out if you wanna hang and talk about sneaky LLM API hacks, accountability, and the geometry of LLM representations!
Wanna know gpt-3.5-turbo's embed size? We find a way to extract info from LLM APIs and estimate gpt-3.5-turbo’s embed size to be 4096. With the same trick we also develop 25x faster logprob extraction, audits for LLM APIs, and more! 📄 arxiv.org/abs/2403.09539 Here’s how 1/🧵
Congratulations to the GDM @GoogleDeepMind team on their best paper award at #ICML2024 & Appreciate @afedercooper's shout out to our concurrent paper 🙌 If you are into the topic of recovering model info through just its output logits, check out our paper led by @mattf1n too!…
Wanna know gpt-3.5-turbo's embed size? We find a way to extract info from LLM APIs and estimate gpt-3.5-turbo’s embed size to be 4096. With the same trick we also develop 25x faster logprob extraction, audits for LLM APIs, and more! 📄 arxiv.org/abs/2403.09539 Here’s how 1/🧵