Victor Veitch's Avatar

Victor Veitch

@vveitch

machine learning and artificial intelligence | University of Chicago / Google

549
Followers 72
Following 5
Posts 13.10.2023
Joined
Posts Following

Latest posts by Victor Veitch @vveitch

come learn about LLM geometry!

24.04.2025 19:48 👍 1 🔁 0 💬 0 📌 0

I'll present this poster tonight at East exhibit hall a-c 2510. 5-7:30 pm.

Come chat about alignment!

12.12.2024 18:47 👍 7 🔁 0 💬 0 📌 0

I'll be at NeurIPS Thursday-Sunday; send me an email if you'd like to chat :)

10.12.2024 02:11 👍 2 🔁 0 💬 0 📌 0
On Spurious Associations and LLM Alignment Large language models are `aligned' to bias them towards outputting responses that are good on various measures---e.g., we may want them to be helpful, factual, and polite. Often, alignment procedures...

in talk form simons.berkeley.edu/talks/victor...

23.11.2024 23:21 👍 4 🔁 0 💬 0 📌 0
On Spurious Associations and LLM Alignment Large language models are `aligned' to bias them towards outputting responses that are good on various measures---e.g., we may want them to be helpful, factual, and polite. Often, alignment procedures...

LLM Alignment aims at making model outputs preferred by a ranker while changing as little 'off-target' behavior as possible.

Turns out:
-best-of-$n$ is the optimal option!
-you can contrastively train an LLM to mimic its own best-of-$n$ distribution!

BonBon alignment: arxiv.org/abs/2406.00832

23.11.2024 23:21 👍 6 🔁 0 💬 1 📌 1