#Paper_Files
LightNews LightNews
paulhansel.com
You're not going to believe this

proceedings.neurips.cc/paper_files/...
Scaling transformer neural networks for skillful and reliable medium-range weather forecasting
proceedings.neurips.cc
August 20, 2026 at 2:18 AM
astronglitch.bsky.social
arxiv.org/abs/2501.05803
arxiv.org/abs/2406.14272
proceedings.neurips.cc/paper_files/...
(13/23)
March 30, 2025 at 3:27 PM
timnitgebru.blacksky.app
-Scaling laws? Some people looked at some plots and came up with that.

-"Emergence"? Take a look at this paper showing how that is nothing but hot air: proceedings.neurips.cc/paper_files/...
April 10, 2025 at 11:29 PM
abeba.blacksky.app
no. feel welcome to read my work on dataset audit
1. arxiv.org/pdf/2110.01963
2. ieeexplore.ieee.org/abstract/doc...
3. proceedings.neurips.cc/paper_files/...
4. dl.acm.org/doi/pdf/10.1...
January 4, 2025 at 12:55 PM
justincormack.bsky.social
I mean, there are papers on how to solve the arithmetic problem with different numerical embeddings proceedings.neurips.cc/paper_files/...
proceedings.neurips.cc
September 30, 2026 at 6:37 PM
sueyeonchung.bsky.social
*Neuro-inspired & efficient AI:
- MMCR (Maximum Manifold Capacity Representations): proceedings.neurips.cc/paper_files/...
- Contrast-equivariant SSL improves neural-model alignment
proceedings.neurips.cc/paper_files/...

2/3
June 10, 2025 at 12:15 AM
e-mum.bsky.social
Some papers related to this have been linked such as proceedings.iclr.cc/paper_files/... and arxiv.org/abs/2502.00873
Arithmetic Without Algorithms: Language Models Solve Math with a Bag of Heuristics
proceedings.iclr.cc
October 2, 2026 at 7:12 PM
shahabbakht.bsky.social
I’m thinking even beyond image labeling—more in the realm of recent self-supervised learning methods (e.g., what we proposed here: proceedings.neurips.cc/paper_files/...).

There’s really cool work bringing these concepts into visual neuroscience (we and others have been doing it),…
seq-JEPA: Autoregressive Predictive Learning of Invariant-Equivariant World Models
proceedings.neurips.cc
October 4, 2026 at 1:10 AM
sjgreenwood.bsky.social
As such, we’re excited about this opportunity to put into practice the research ideas around individual and societal recommendation quality that our team has been thinking about (proceedings.neurips.cc/paper_files/..., dl.acm.org/doi/10.1145/...)
March 10, 2025 at 3:12 PM
astronglitch.bsky.social
These numbers are basically impossible even for Disney-sized companies to create fully on their proprietary data, let alone a small company or individual.
github.com/CompVis/stab...
laion.ai/blog/laion-5b/
proceedings.neurips.cc/paper_files/...
(19/23)
March 30, 2025 at 3:27 PM
lnalegre.bsky.social
It is really cool to see our work on multi-step GPI being cited in this amazing survey! :)

proceedings.neurips.cc/paper_files/...
September 3, 2025 at 12:44 PM
chriswolfvision.bsky.social
TIL that Seldonian RL is a thing, like in "Hari Seldon", the guy in Isaac Asimov's "Foundation" who predicted the future of an entire galaxy over the span of 1000 years (now _THAT_ is a capable internal world model!).

Screenshot from papers.nips.cc/paper_files/...
June 15, 2025 at 8:03 AM
stefanabaumann.bsky.social
Hasn't this idea been around for a while? E.g., proceedings.neurips.cc/paper_files/...
proceedings.neurips.cc
October 31, 2025 at 7:49 PM
pcastr.bsky.social
3. Sutton et al., “Policy gradient methods for reinforcement learning with function approximation”
The policy gradient theorem is another cornerstone of RL research, we should all be able to replicate this proof!

papers.nips.cc/paper_files/...
December 6, 2024 at 8:56 PM
neuranna.bsky.social
I like this one proceedings.neurips.cc/paper_files/...
The Quantization Model of Neural Scaling
proceedings.neurips.cc
November 29, 2025 at 5:38 PM
elisennesh.bsky.social
The virgin finely-tuned irreversible gradient-driven bespoke Sequential Monte Carlo algorithm

vs

The Chad "just add Gaussian noise backwards" proceedings.neurips.cc/paper_files/...
Sampling from multi-modal distributions with polynomial query complexity in fixed dimension via reverse diffusion
proceedings.neurips.cc
September 5, 2026 at 1:37 AM
sueyeonchung.bsky.social
*Interpretable theory of neural-AI alignment:
- Neural prediction: proceedings.neurips.cc/paper_files/...
- Representation similarity:
arxiv.org/pdf/2502.19648

3/3
June 10, 2025 at 12:15 AM
shahabbakht.bsky.social
I like the title of the previous version of their paper better.

proceedings.neurips.cc/paper_files/...
Performance-optimized deep neural networks are evolving into worse models of inferotemporal visual cortex
proceedings.neurips.cc
May 4, 2025 at 12:53 PM
tyrellturing.bsky.social
Related to this earlier, wonderful paper from Joao Sacramento's group (this was the paper that first got me thinking in this direction):

proceedings.neurips.cc/paper_files/...
The least-control principle for local learning at equilibrium
proceedings.neurips.cc
November 13, 2024 at 7:19 PM
adam-lg.bsky.social
25. The ideal of comparing how similar models are in a latent space, a la this paper: proceedings.neurips.cc/paper_files/..., is fascinating, and I wish more people explored this idea
proceedings.neurips.cc
July 27, 2025 at 5:30 PM
chazfirestone.bsky.social
And here’s the new work he presented today!

proceedings.neurips.cc/paper_files/...

proceedings.neurips.cc/paper_files/...

👀
Break It Down: Evidence for Structural Compositionality in Neural Networks
proceedings.neurips.cc
March 31, 2025 at 9:58 PM
spmontecarlo.bsky.social
I got back onto this topic via arxiv.org/abs/2312.05910; an early paper is proceedings.mlr.press/v9/turner10a..., and a couple of intermediate works which I found useful are proceedings.neurips.cc/paper_files/... and proceedings.mlr.press/v97/ialongo1....
October 12, 2025 at 6:18 PM
e-mum.bsky.social
I am not sure that is correct. It could also be further evidence (I am agnostic about it). proceedings.iclr.cc/paper_files/...
Arithmetic Without Algorithms: Language Models Solve Math with a Bag of Heuristics
proceedings.iclr.cc
September 30, 2026 at 4:40 PM
david-klindt.bsky.social
Goodhart's law; Kept on pushing, unsuccessfully, to have that mentioned in here proceedings.neurips.cc/paper_files/...
Measuring Per-Unit Interpretability at Scale Without Humans
proceedings.neurips.cc
March 25, 2025 at 6:03 PM
jeanremiking.bsky.social
Self supervised learning accounts for the brain hierarchy of speech representations :
Eddie Chang's latest study just out in Nature Neuro: www.nature.com/articles/s41...

Only grumpy comment: Given its similarity, I do wish our study had been cited/discussed: proceedings.neurips.cc/paper_files/...
October 31, 2023 at 9:00 AM