Devin White
@DevinWhiteAI
Followers
18
Following
186
Media
14
Statuses
63
ML Researcher @USAEOP. Pushing RLHF forward & using LLMs to master gameplay.
Joined February 2024
Thank you to everyone who stopped by our presentations yesterday! I had a great time sharing our work and chatting with so many of you.
0
0
0
If you're at #ICML2025 🇨🇦, join us today for our "Too Big to Think" oral presentation at 9:30AM (Room 215-216) and "Multi-Task Reward Learning from Human Ratings" poster at 12PM (Ballroom A)! See you there! #TinyTitans #RLHF
0
0
0
Multi-Task Reward Learning from Human Ratings.
arxiv.org
Reinforcement learning from human feedback (RLHF) has become a key factor in aligning model behavior with users' goals. However, while humans integrate multiple strategies when making decisions,...
0
1
2
Paper - https://t.co/I4WG0jEffJ Paper Title: "Too Big to Think: Capacity, Memorization, and Generalization in Pre-Trained Transformers"
0
1
4
Big news! 🎉 Our paper “Multi-Task Reward Learning from Human Ratings” was accepted to the Models of Human Feedback for AI Alignment workshop at #ICML2025! In this paper we treat ratings not just as class labels, but as rich reward signals with underlying structure and scale.
0
1
3
🚨Exciting news!🚨 Our paper, “Too Big to Think: Capacity, Memorization, and Generalization in Pre-Trained Transformers”, was accepted for an oral presentation at the Tiny Titans: The next wave of On-Device Learning for Foundational Models workshop (@tinytitans_icml) at
0
1
3
🚨 Big news! Simple RbRL now has a sleek, user-friendly interface! 🖥️✨ You can now: ✅ Rate trajectories directly in the UI ✅ Train RL agents with human feedback ✅ Explore Rating-based RL hands-on with ease This lightweight, open-source tool makes RbRL accessible to
github.com
Simplified, modern implementation of Rating and Preference-based Reinforcement Learning. - Dev1nW/Simplified-Rating-and-Preference-RL
0
0
3
OpenAI just dropped a GitHub connector for ChatGPT’s Deep Research Now you can plug into GitHub repos to search code, scan PRs, and auto-generate detailed, citation-backed reports — all inside ChatGPT. Dev workflows just got smarter https://t.co/8nPwMGzI0n
4
22
174
🚀 Big update to Atari-GPT! ✨ Progress bar during testing (steps & reward) ✨ Cleaner function definitions for ease of use ✨ Easy game/model selection via CLI ✨ New analysis file to visualize results Perfect for Atari AI fans! Try it out and share your results!
0
0
2
Check out all the things we have been working on here:
scholar.google.com
Machine Learning Researcher, Army Educational Outreach Program - Cited by 44 - RLHF - Human Guided Reinforcement Learning - AI Alignment - Large Language Models - Small Language Model
0
0
1
🎉I’m excited to say that I have reached a small but personal milestone of 20 citations! I want to say a huge thank you to everyone who I have had the honor of collaborating with and I'm excited for what's next!
1
0
3
Learning from Negative Feedback, Positive Feedback, or Both: https://t.co/5g8r7mtg1F RbRL2.0: Integrated Reward and Policy Learning for Rating-based Reinforcement Learning:
arxiv.org
Reinforcement learning (RL), a common tool in decision making, learns policies from various experiences based on the associated cumulative return/rewards without treating them differently. On the...
0
0
1
Had a blast presenting Atari-GPT at the Toward Knowledgeable Foundation Models Workshop @RealAAAI! Check out the full paper here: https://t.co/JN8uCxyl40
#AI #AAAI2025 #LLMs
0
0
3
Today is the day! Welcome to our 2nd workshop on Knowledgeable Foundation Models in Room 112. Come and talk with these wonderful speakers @ehovy @Wenpeng_Yin @RICEric22 @Lianhuiq @liharryzhang @HuajieShaoML ! Special thanks to our organizers @ZoeyLi20 @megamor2 @XiaozhiWangNLP
2
19
72
Curious how Atari-GPT blends Atari's retro feel with advanced LLMs? Discover the magic here:
0
1
4
Enhanced collision detection in my ASCII Breakout game to test GPT-4o, with this it got better results than ever! See the code and try it out: https://t.co/R0pHmkPWHd
#AI #LLM
0
0
3
Exciting update on my Simple RLHF codebase! New results replicate the original RbRL paper, but cut training time by ~33% and run smoothly on modern hardware (like M-series Apple Silicon). Curious? Dive into the details here: https://t.co/B03QWQG0LQ
#RLHF #AI #MachineLearning
1
0
5