When reviewing, I will give credit to submissions which work well in an elegantly simple pipeline. While other reviewers (always) raise novelty issues (and/or out-of-the-scope questions for the specific task). Look forward to seeing how industrial research looks like.
INNOVATION
-
Llama.cpp Introduces Full CUDA GPU Acceleration Support
By
–
Llama.cpp: Full CUDA GPU Acceleration PR: https://
github.com/ggerganov/llam
a.cpp/pull/1827
… -

Transformers Learn Through Gradual Rank Increase
By
–
Transformers learn through gradual rank increase paper page: https://
huggingface.co/papers/2306.07
042
… identify incremental learning dynamics in transformers, where the difference between trained and initial weights progressively increases in rank. We rigorously prove this occurs under the -

Cap3D: Scalable Automatic 3D Object Text Description System
By
–
Scalable 3D Captioning with Pretrained Models paper page: https://
huggingface.co/papers/2306.07
279
…
dataset: https://
huggingface.co/datasets/tiang
e/Cap3D
… introduce Cap3D, an automatic approach for generating descriptive text for 3D objects. This approach utilizes pretrained models from image captioning, image-text -

Retrieval-Enhanced Contrastive Vision-Text Models for Improved Concept Recognition
By
–
Retrieval-Enhanced Contrastive Vision-Text Models paper page: https://
huggingface.co/papers/2306.07
196
… Contrastive image-text models such as CLIP form the building blocks of many state-of-the-art systems. While they excel at recognizing common generic concepts, they still struggle on -

FasterViT: Fast Vision Transformers with Hierarchical Attention
By
–
FasterViT: Fast Vision Transformers with Hierarchical Attention paper page: https://
huggingface.co/papers/2306.06
189
… design a new family of hybrid CNN-ViT neural networks, named FasterViT, with a focus on high image throughput for computer vision (CV) applications. FasterViT combines the -

Benchmarking Neural Network Training Algorithms: Performance Analysis
By
–
Benchmarking Neural Network Training Algorithms paper page: https://
huggingface.co/papers/2306.07
179
… Training algorithms, broadly construed, are an essential part of every deep learning pipeline. Training algorithm improvements that speed up training across a wide variety of workloads (e.g., -

Controlling Text-to-Image Diffusion by Orthogonal Finetuning
By
–
Controlling Text-to-Image Diffusion by Orthogonal Finetuning paper page: https://
huggingface.co/papers/2306.07
280
… Large text-to-image diffusion models have impressive capabilities in generating photorealistic images from text prompts. How to effectively guide or control these powerful models to -

Face0: Instant Face Conditioning for Text-to-Image Models
By
–
Face0: Instantaneously Conditioning a Text-to-Image Model on a Face paper page: https://
huggingface.co/papers/2306.06
638
… present Face0, a novel way to instantaneously condition a text-to-image generation model on a face, in sample time, without any optimization procedures such as fine-tuning or -
Schmidhuber’s 1991 Alternative to RNNs: The Origins of Linear Transformers
By
–
Hah, yeah, little known fun fact: Schmidhuber proposed an alternative to RNNs back in 1991, which is now called "linear Transformers" or "Transformers with linearized self-attention" via more recent papers. Summarized it here: https://
magazine.sebastianraschka.com/p/why-the-orig
inal-transformer-figure
…