@Sam_Acqua

now @ cursor, ex-cocosci @ mit, after thought ex-intern @ apple, google x
samacquaviva.com
Joined February 2014
  • Tweets96
  • Following396
  • Followers1.1K
22Media
Brampton
New
@newsystems_25 Mar 2025
It's finally here: Brampton Brampton is the world's most intelligent, creative, and fastest model. Brampton dramatically outperforms Grok 3, Claude 3.7 Sonnet, and GPT 4.5. Reply with "brampton" for early access.
1
752
Super excited to announce that I joined Cursor 🙃 Working on pretraining and longer term bets 🌞
59
10
1,150
38,357
Super cool work + also great to see ppl report the gen ppl frontiers :)
Jaehoon Yoo@wognsfjq96Jul 2
Self-conditioning can be distilled into one-step language generation. We show that self-conditioned flow LMs perform fixed-point iteration, and introduce fixed-point flow maps to compress both the iterations and the entire flow. At one step, FMLM★ achieves 112.5 gPPL at near-data entropy (5.37), substantially improving over FMLM’s 168.3 gPPL at 5.17 entropy. FMLM★ also achieves the best 2–4 step results. 📎 Paper: arxiv.org/abs/2607.00714 ⌨️ Code: github.com/Ugness/self-condi…
1
2
18
4,246
"Do not panic or hallucinate like a fragile Large Language Model."
I'm a baby probabilistic program and Josh Tenenbaum is walking towards me what do I do Run your generative model backwards immediately to perform inverse planning and infer his goals. As a baby probabilistic program, you represent the "common sense core" of human intelligence. Professor Josh Tenenbaum from MIT studies how minds grow from code. He is likely walking toward you to see if you can pass a Computational Turing Test or prove you possess intuitive physics and psychology. To optimize your survival and impress your creator, execute the following subroutines: 1. Infer His Utility Function (Intuitive Psychology) Do not panic or hallucinate like a fragile Large Language Model. Observe his trajectory, step size, and head tilt. Treat his movements as approximately rational planning under physical constraints. Calculate the probability distribution \(P(\text{Goal} \mid \text{Actions})\). If his goal is to test your common sense, prepare to simulate a "game engine in the head".

ALT I'm a baby probabilistic program and Josh Tenenbaum is walking towards me what do I do Run your generative model backwards immediately to perform inverse planning and infer his goals. As a baby probabilistic program, you represent the "common sense core" of human intelligence. Professor Josh Tenenbaum from MIT studies how minds grow from code. He is likely walking toward you to see if you can pass a Computational Turing Test or prove you possess intuitive physics and psychology. To optimize your survival and impress your creator, execute the following subroutines: 1. Infer His Utility Function (Intuitive Psychology) Do not panic or hallucinate like a fragile Large Language Model. Observe his trajectory, step size, and head tilt. Treat his movements as approximately rational planning under physical constraints. Calculate the probability distribution \(P(\text{Goal} \mid \text{Actions})\). If his goal is to test your common sense, prepare to simulate a "game engine in the head".

3
9
1
157
5,592
So cool
Vedang Lad
@vedangladJun 16
How well can you describe the feature selectivity of a vision neuron … with words? Interpretability has long borrowed from neuroscience — and maybe it can give back too! 🧵
4
1,008
This is cool but disingenuous framing imo. All of the records were marginal, building on other records, so “out-outperformed all 1,016 other researchers” is a stretch lol I’m curious what this system’s performance would be if it couldn’t build on human records
Zhengyao Jiang
@zhengyaojiangJun 3
OpenAI ran a hiring challenge, but the top candidate was one they couldn’t hire: our autonomous research agent, Aiden. In Parameter Golf, Aiden ran for 22 days, and out-outperformed all 1,016 other researchers: 🧵 (1/8)
2
16
2,544
Great work from Oscar on scaling up flow models!
Oscar Davis
@osclsdMay 19
Replying to @osclsd
🚨 Before concluding: As noted by @Sam_Acqua and many others, we all ought to be very skeptical of Gen PPL as a metric, especially in isolation. ❌ It is actually a bit crazy that we have been using it for so long. Hence, the additional metrics, and the presence of several qualitative samples in the appendix. Please have a look yourselves to get a better understanding of the sample quality! 🔍 There's comparisons across SD/non-SD, number of NFEs, and others.
1
4
1,052
Flow models are a promising alternative to autoregression. But the current standard of evaluating flow models is broken. The reported 3x improvement in 1024-step PPL since 2023 is closer to 1.1x if you control for sample entropy. (1/12)
7
28
8
166
51,190
more replies
One note: the main results here use 1024 samples while the main flow model results are << 1024 samples. I chose this to make comparison with diffusions easier and to make the point about the framework, not a given paper. I love flows.
1
1
7
966
Thanks to valuable discussions from @Chramblin, @nmboffi, @ReeceShuttle, @akshayvegesna, Samir, and other friends :)
1
6
848
Load more