matt (@MattVMacfarlane) / X

matt

339 posts

matt

@MattVMacfarlane

Working on World Models ,PhD @AmlabUvA prev @Microsoft. Living @mission__ctrl, Views Are Not My Own 🏴󠁧󠁢󠁳󠁣󠁴󠁿

Amsterdam

Joined September 2024

Pinned
matt
@MattVMacfarlane
Nov 15, 2024
Check out our work on the Latent Program Network for inductive program synthesis! It is a new architecture for latent program search that enables efficient test-time adaptation without the need for parameter fine-tuning.
Clem Bonnet
@ClementBonnet16
Nov 15, 2024
Introducing Latent Program Network (LPN), a new architecture for inductive program synthesis that builds in test-time adaption by learning a latent space that can be used for search 🔎 Inspired by @arcprize 🧩, we designed LPN to tackle out-of-distribution reasoning tasks!
14K
matt
@MattVMacfarlane
Jan 22, 2025
Replying to @deedydas
Jane Street is really starting to rack up some notorious alumni
27K
matt
@MattVMacfarlane
Jan 12, 2025
1/ An interesting outcome of the @arcprize 2024 was the high performance of test-time fine-tuning. It is common in the field of AI to emphasize novelty over historical connections, so I'd like to highlight an early application of this exact idea, applied to program synthesis from
8.7K
matt
@MattVMacfarlane
Dec 6, 2024
Great to have Searching Latent Program Spaces (arxiv.org/abs/2411.08706) recognised with a 🥉3rd place paper award in the @arcprize! It was a pleasure working on this with @ClementBonnet16 . Looking forward to continuing to develop these ideas even further.
François Chollet
@fchollet
Dec 6, 2024
Today we're announcing the winners of ARC Prize 2024. We're also publishing an extensive technical report on what we learned from the competition (link in the next tweet). The state-of-the-art went from 33% to 55.5%, the largest single-year increase we've seen since 2020. The
arxiv.org
Searching Latent Program Spaces
General intelligence requires systems that acquire new skills efficiently and generalize beyond their training distributions. Although program synthesis approaches have strong generalization...
2.6K
matt
@MattVMacfarlane
Jan 9, 2025
Replying to @polynoamial
Citing this would essentially just be citing a black box. Everything about this model is closed source, including the method. Are we just supposed to trust it was trained using some vague Reinforcement Learning and go ahead and add it to the literature?
959
matt
@MattVMacfarlane
Dec 8, 2024
Replying to @giffmana and @_aidan_clark_
Arguably RL for games brought in a lot of money and attention to AI, so maybe it was unnecessary from a research point of view but it probably was necessary practically.
2.1K
matt
@MattVMacfarlane
Feb 18, 2025
Checkout @ClementBonnet16 discussing searching latent program spaces tommorow on @MLStreetTalk !
Machine Learning Street Talk
@MLStreetTalk
Feb 18, 2025
We spoke with @ClementBonnet16 at NeurIPS about his extremely innovative approach to the @arcprize using a form of test time inference where you search a latent space of a VAE before making an optimal prediction. @fchollet was so impressed, he hired Clem shortly after! 😃 -
00:00
2K
matt
@MattVMacfarlane
Dec 10, 2024
If you are interested in scalable search based 🔎 policy improvement operators come chat with me at our poster on SPO tommorow!
InstaDeep
@instadeepai
Dec 10, 2024
Excited to share our latest work on Sequential Monte Carlo Policy Optimisation (SPO)🔥— a scalable, search-based RL algorithm leveraging SMC as a policy improvement operator for both continuous and discrete environments! 📍 Catch us tomorrow at #NeurIPS2024 (poster #94776) from
GIF
1.3K
matt
@MattVMacfarlane
Jan 12, 2025
Replying to @MattVMacfarlane
3/ Searching Latent Program Spaces @ClementBonnet16 (arxiv.org/abs/2411.08706) similarly uses a meta-induction network, but our insight was that gradient descent can be performed in the compact task embedding space, as opposed to the parameter space.
arxiv.org
Searching Latent Program Spaces
General intelligence requires systems that acquire new skills efficiently and generalize beyond their training distributions. Although program synthesis approaches have strong generalization...
865
matt
@MattVMacfarlane
Dec 6, 2024
Excited to share I’ll be at #NeurIPS2024 🔥I will be presenting SPO: Sequential Monte Carlo Policy Optimisation: arxiv.org/pdf/2402.07963 🔎. If you'd like to discuss this or topics like search, test-time adaptation, or program synthesis reach out and we can catch up in person.
603
matt
@MattVMacfarlane
Dec 26, 2024
Replying to @polynoamial @fchollet and @mikeknoop
to be fair ARC-AGI wasn't beaten. There are compute requirements for the competition.
644
matt
@MattVMacfarlane
Nov 18, 2024
Replying to @y0b1byte
Thanks for sharing! Here's the original post:
Clem Bonnet
@ClementBonnet16
Nov 15, 2024
Introducing Latent Program Network (LPN), a new architecture for inductive program synthesis that builds in test-time adaption by learning a latent space that can be used for search 🔎 Inspired by @arcprize 🧩, we designed LPN to tackle out-of-distribution reasoning tasks!
1.9K
matt
@MattVMacfarlane
Dec 25, 2024
Epic Xmas haul ⛄ looking forward to getting stuck in #activeinference @NickJChater @MITCoCoSci @ClementBonnet16
724
matt
@MattVMacfarlane
Feb 28, 2025
Replying to @MLStreetTalk
This makes me more bullish on the benefits of Mixture of Experts models and reducing the number of active parameters during inference 💪
1.4K