Patrick Chao
@patrickrchao
Followers
942
Following
417
Media
14
Statuses
53
research @openai, making models smarter and safer
Joined January 2014
if you think gpt-4.5 is big wait til you see gpt-4.11
GPT-4.5 has entered the Chat. https://t.co/tBzJxSyCeY
101
124
5K
This is one of the craziest graphs I've ever seen! AI Models went from dragging humans down (gpt-4o) → to breaking past the human baseline gpt-5 delivers ~1.6× efficiency in both speed and cost 📈
Today we’re introducing GDPval, a new evaluation that measures AI on real-world, economically valuable tasks. Evals ground progress in evidence instead of speculation and help track how AI improves at the kind of work that matters most. https://t.co/uKPPDldVNS
0
1
7
My team forced me to post this. It's definitely the last time I organize a nice offsite for them 😤😂
22
3
203
huge congrats to @_chris_lu_ @minyoung_huh and @SuvanshSanjeev for putting a truly herculean effort into training gpt-5 mini and nano!
GPT-5 is what you’ve been waiting for – it defines and extends the cost-intelligence frontier across model sizes today. it’s been a long journey, and we’ve landed pivotal improvements across many axes in the whole GPT-5 family. and hey no more model picker (by default)!
1
1
14
After rejections at ICLR, ICML, and NeurIPS, I'm happy to report that "Jailbreaking Black Box LLMs in Twenty Queries" (i.e., the PAIR paper) has been accepted at @satml_conf! 🚀 A quick 🧵 summarizing some thoughts a year on from PAIR's release.
6
19
175
What happens when you train a video generation model to be conditioned on motion? Turns out you can perform "motion prompting," just like you might prompt an LLM! Doing so enables many different capabilities. Here’s a few examples – check out this thread 🧵 for more results!
20
145
672
All Plus and Team users in ChatGPT
Advanced Voice is rolling out to all Plus and Team users in the ChatGPT app over the course of the week. While you’ve been patiently waiting, we’ve added Custom Instructions, Memory, five new voices, and improved accents. It can also say “Sorry I’m late” in over 50 languages.
207
219
4K
models are kinda good check out our system card for some cool evals and plots!
openai.com
This report outlines the safety work carried out prior to releasing OpenAI o1 and o1-mini, including external red teaming and frontier risk evaluations according to our Preparedness Framework.
We're releasing a preview of OpenAI o1—a new series of AI models designed to spend more time thinking before they respond. These models can reason through complex tasks and solve harder problems than previous models in science, coding, and math.
1
2
38
it’s a good model sir
We're releasing a preview of OpenAI o1—a new series of AI models designed to spend more time thinking before they respond. These models can reason through complex tasks and solve harder problems than previous models in science, coding, and math.
6
4
311
We’re sharing the GPT-4o System Card, an end-to-end safety assessment that outlines what we’ve done to track and address safety challenges, including frontier model risks in accordance with our Preparedness Framework.
openai.com
This report outlines the safety work carried out prior to releasing GPT-4o including external red teaming, frontier risk evaluations according to our Preparedness Framework, and an overview of the...
215
333
2K
Some folks from OpenAI’s Preparedness and agent safety efforts will be at the AI Security Forum and Defcon in Las Vegas next week. If you’d like to chat with us there about AI + cybersecurity, fill out this form (by Aug 5)! https://t.co/aab5F24t8g
docs.google.com
Some members of OpenAI's Preparedness and Agent Safety teams will be at the AI Security Forum and Defcon this year. Fill out this form if you'd like to chat with us! It is best to fill out this form...
1
8
54
🚨 We are very excited to release JailbreakBench v1.0! 📄 We have substantially extended the version 0.1 that was on arXiv since March: - More attack artifacts (Prompt template with random search in addition to GCG, PAIR, and JailbreakChat): https://t.co/WCLixoM3fO. - More
3
26
115
Great to see that both of our recent papers—JailbreakBench ( https://t.co/WgzTj56PKB) and our adaptive attack paper ( https://t.co/nlebNR6GB5)—have been used by Google to evaluate the robustness of Gemini 1.5 Flash/Pro against jailbreaking attacks! An interesting comment from
2
15
115
What do you see in these images? These are called hybrid images, originally proposed by Aude Oliva et al. They change appearance depending on size or viewing distance, and are just one kind of perceptual illusion that our method, Factorized Diffusion, can make.
10
102
449
For more details, check out: Paper: https://t.co/bUNqVRCUo8 Code: https://t.co/o17WWX8c1B Leaderboard: https://t.co/sWgg11C4W7 This could not be possible without our great collaborators: @edoardo_debe, @maksym_andr, @AlexRobey23, @fra__31, @VSehwag_, @EdgarDobriban, @tml_lab,
github.com
JailbreakBench: An Open Robustness Benchmark for Jailbreaking Language Models [NeurIPS 2024 Datasets and Benchmarks Track] - JailbreakBench/jailbreakbench
0
3
12
At the moment, we have added the following attacks and defenses (more are coming soon). Even on Vicuna-13B, the attack success rate is far from 100% - there is a lot of room for improvement! Consider submitting your own attacks/defenses. 🧵7/n
1
0
3
Submitting is easy! You can create your submission to our leaderboard in just three lines of Python: 🧵6/n
1
0
3