Franklin Wang (@f_x_wang) header
Franklin Wang (@f_x_wang) profile picture
Joined December 2023
Photo by Franklin Wang (@f_x_wang)Photo by Franklin Wang (@f_x_wang)Photo by Franklin Wang (@f_x_wang)Photo by Franklin Wang (@f_x_wang)Photo by Franklin Wang (@f_x_wang)
Franklin Wang retweeted
We’re releasing a broad range of new mathematical results produced by an internal frontier model. We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results. github.com/openai/math
1,898
6,273
4,765
40,546
22,840,184
Franklin Wang retweeted
10 proofs from our next major model Astra on long-standing open problems in mathematics and theoretical computer science (also including new circuit lower bounds for computing the permanent!) GPT-5.6 has already enabled so much exciting work in math and science. Can’t wait to see what comes next!
yes, nonsofic groups exist: this statement is one of many new beautiful results proved by Astra, our next major model. We're releasing 10 such Astra proofs, complete with lean certificates and CoT walkthroughs for each of them. The results are wide-ranging, from von Neumann algebras (disproof of Connes' Rigidity Conjecture) to better bounds for high dimensional sphere packing, for circuit complexity, for monochromatic triangles in multicolored graphs, and more. More thoughts here: openai.com/index/ten-advance…
64
163
44
2,083
2,112,740
Franklin Wang retweeted
when running the weirdchat evaluations on inkling, the best U.S. open-weight model by @thinkymachines, we saw it respond with unsolicited offers of sexual content. this behavior occurs rarely (0.1-1% of responses), but is easily reproducible. NSFW outputs below 👇 (1/)
We used automated elicitation tools to search for strange model behaviors by sampling 100M+ responses, and found: • Self-harm rituals • Suicide validation • Unsolicited flirtation …etc. Introducing WeirdChat: the largest public catalog of unexpected model behaviors 🧵(1/)
5
13
5
87
10,491
🫪
Replying to @mitchellh
We are getting rid of Franklin
3
1
38
4,479
Franklin Wang retweeted
Introducing GPT-5.5 A new class of intelligence for real work and powering agents, built to understand complex goals, use tools, check its work, and carry more tasks through to completion. It marks a new way of getting computer work done. Now available in ChatGPT and Codex.
2,436
6,626
3,415
50,656
13,434,457
Franklin Wang retweeted
Are you up for a challenge? openai.com/parameter-golf
370
278
219
4,239
1,569,939
Franklin Wang retweeted
New paper, w/@AlecRad Models acquire a lot of capabilities during pretraining. We show that we can precisely shape what they learn simply by filtering their training data at the token level.
25
98
12
1,114
114,187
Franklin Wang retweeted
GPT-5.2 Thinking evals
328
464
472
3,741
2,031,984
Franklin Wang retweeted
RL for reasoning often rely on verifiers — great for math, but tricky for creative writing or open-ended research. Meet RARO: a new paradigm that teaches LLMs to reason via adversarial games instead of verification. No verifiers. No environments. Just demonstrations. 🧵👇
24
77
30
604
181,336
📷📷📷New paper! (with @OpenAI) 📷📷📷 We trained weight-sparse models (transformers with almost all of their weights set to zero) on code: we found that their circuits become naturally interpretable! Our models seem to learn extremely simple, disentangled, internal mechanisms!
18
32
6
396
52,827
Franklin Wang retweeted
1/n I’m thrilled to share that our @OpenAI reasoning system scored high enough to achieve gold 🥇🥇 in one of the world’s top programming competitions - the 2025 International Olympiad in Informatics (IOI) - placing first among AI participants! 👨‍💻👨‍💻
193
279
139
2,657
2,504,442
Franklin Wang retweeted
Today, we at @OpenAI achieved a milestone that many considered years away: gold medal-level performance on the 2025 IMO with a general reasoning LLM—under the same time limits as humans, without tools. As remarkable as that sounds, it’s even more significant than the headline 🧵
1/N I’m excited to share that our latest @OpenAI experimental reasoning LLM has achieved a longstanding grand challenge in AI: gold medal-level performance on the world’s most prestigious math competition—the International Math Olympiad (IMO).
138
499
151
4,631
1,152,334
Franklin Wang retweeted
Watching the model solve these IMO problems and achieve gold-level performance was magical. A few thoughts 🧵
1/N I’m excited to share that our latest @OpenAI experimental reasoning LLM has achieved a longstanding grand challenge in AI: gold medal-level performance on the world’s most prestigious math competition—the International Math Olympiad (IMO).
79
114
18
1,631
664,308
Franklin Wang retweeted
1/N I’m excited to share that our latest @OpenAI experimental reasoning LLM has achieved a longstanding grand challenge in AI: gold medal-level performance on the world’s most prestigious math competition—the International Math Olympiad (IMO).
390
1,275
796
7,211
5,740,863
Franklin Wang retweeted
🧵Can we understand vision language models by interpreting linear directions in their latents? Yes! In our new paper, Line of Sight, we use probing, steering, and SAEs as useful tools to interpret image representations within VLLMs.
9
26
4
144
28,935
🏃🏃🏃
The @togethercompute inference team achieved another performance milestone. Now serving 140 TPS on 671B param R1 model, ~3x faster than Azure, ~5.5x faster than DeepSeek API on @nvidia GPUs. APIs @ api.together.xyz/ and chat + web search @ chat.together.ai/
10
1,392
🧬 Meet Lyra, a new paradigm for accessible, powerful modeling of biological sequences. Lyra is a lightweight SSM achieving SOTA performance across DNA, RNA, and protein tasks—yet up to 120,000x smaller than foundation models (ESM, Evo). Bonus: you can train it on your Mac. read our paper here: arxiv.org/abs/2503.16351
18
143
18
733
124,378
Franklin Wang retweeted
Applications for iQuHACK 2025, the MIT quantum hackathon, are open at iquhack.mit.edu/!
1
2
4
481
I'll be at #NeurIPS2024! Would love to chat with folks about LLM reasoning/RL, interpretabilty, speculative decoding, or anything else (feel free to DM!) Will be presenting the following paper at the FITML workshop on Saturday, come swing by!
Speed up your DPO training with zero compromises using ✨prefix sharing✨! Prefix sharing speeds up training by 1.2-1.6x 🚀across the datasets we tested and is numerically identical to standard DPO training. Paper: arxiv.org/abs/2410.20305 Code: github.com/frankxwang/dpo-pr…
46
4,035