Ben Hoover

@Ben_Hoov

Building the next generation of energy models. @prefrontal_ai

London

bhoov.com

2018年2月からXを利用しています

422 フォロー中

1,871 フォロワー

固定
Ben Hoover
@Ben_Hoov
11月13日
Big reasoning speedups are here. 1,000× faster training and 280× less compute than any looped model. All done using a fully trainable energy from the 80s Keep reading to hear how energy is the right paradigm for reasoning With @HughBlayney @jcrwhittington @prefrontal_ai 🧵/10
43
Ben Hoover
@Ben_Hoov
11月2日
Pushing the frontiers on long context modeling. All open source. Great work team 🫡
James Whittington
@jcrwhittington
11月2日
Today we are happy to share one of our first open research releases: Latent Sparse Attention (LSA), a step towards subquadratic attention that matches accuracy of full attention at ultra-long context. Sparse attention attends to only a few tokens per query, but it has to find
1
Ben Hoover
@Ben_Hoov
4月20日
Transformer looping = recurrence = naive weight sharing. A simple idea that works great But look deeper. Can the recurrence be solving an optimization problem? This is the goal of understanding transformers as energy-based models. And it unlocks serious connections to physics 👇
Hayden Prairie
@hayden_prairie
4月15日
We’ve been thinking a lot about scaling laws, wondering if there is a more effective way to scale FLOPs without increasing parameters. Turns out the answer is YES – by looping blocks of layers during training. We find that predictable scaling laws exist for layer looping,
2
Ben Hoover
@Ben_Hoov
4月10日
NRGPT shows how causal inference can be viewed as explicit optimization: next-token prediction becomes *energy minimization over token states*. Pretraining shapes the energy. Generation explores it Excited to scale these ideas & bring the physics of memory to foundation models💪
Dmitry Krotov
@DimaKrotov
4月9日
GPT is a standard paradigm for language models: 👉 tokens → neural net → shifted tokens Energy-based models (Hopfield style) is a different paradigm: 👉 tokens evolve on an energy landscape What if you could have both? In our new #ICLR2026 paper, we build a hybrid: •
Ben Hoover
@Ben_Hoov
2月4日
📢 Call for Papers | New Frontiers of Associative Memory (NFAM) @iclr_conf'26 Deadline: Feb 14 AoE (10 days!) Welcoming a range of topics: Hopfield Nets, memory, EBMs, neuroscience, bioplausible architectures, neuromorphic hardware... 🥳 More info: nfam2026.amemory.net/cfp/

Ben Hoover

Ben Hooverさんのポスト