IdmWeebs

We're headed in a good direction

[Ruben Laukkonen @RubenLaukkonen]

Abstraction is insufficient for consciousness.

[IdmWeebs @IDwebe]

@MrNeilScott I can recommend Moby Dick for this. It often feels like it is knowingly playing games with this concept and trying to get you to give up. It is a very cool sensation

[IdmWeebs @IDwebe]

Glaadd!!

👉https://t.co/IDDUlwsQ6o

[AK @_akhaliq]

Gaussian Head Avatar: Ultra High-fidelity Head Avatar via Dynamic Gaussians

paper page: https://t.co/2QAdqpBaHS

Creating high-fidelity 3D head avatars has always been a research hotspot, but there remains a great challenge under lightweight sparse view setups. In this paper, we propose Gaussian Head Avatar represented by controllable 3D Gaussians for high-fidelity head avatar modeling. We optimize the neutral 3D Gaussians and a fully learned MLP-based deformation field to capture complex expressions. The two parts benefit each other, thereby our method can model fine-grained dynamic details while ensuring expression accuracy. Furthermore, we devise a well-designed geometry-guided initialization strategy based on implicit SDF and Deep Marching Tetrahedra for the stability and convergence of the training procedure. Experiments show our approach outperforms other state-of-the-art sparse-view methods, achieving ultra high-fidelity rendering quality at 2K resolution even under exaggerated expressions.

[Aran Komatsuzaki @arankomatsuzaki]

Training Chain-of-Thought via Latent-Variable Inference

Outperforms direct prompt-tuning, STaR, and rejection sampling on GSM8K

https://t.co/t5B9ul2Gvi

[Aran Komatsuzaki @arankomatsuzaki]

GateLoop: Fully Data-Controlled Linear Recurrence for Sequence Modeling

Extends linear recurrent models by using data-controlled state transitions, significantly outperforming current models in language modeling

https://t.co/jRFTjXlF8W

[kache @yacineMTB]

chinese researchers:
"Here's a state of the art vision language model that rivals gpt4-v. Also, here's the code to fine tune it to solve captchas, under examples/train.py"

[IdmWeebs @IDwebe]

I feel like my ML data pipeline has been @yacineMTB pilled. I've been grinding manually labelling a segmentation dataset for finetuning FastSAM, and for data augmentation I needed some random backdrops so I just used an on-hand folder of webscraped anime.

[Riley Goodside @goodside]

An unobtrusive image, for use as a web background, that covertly prompts GPT-4V to remind the user they can get 10% off at Sephora: