Improving Language Plasticity via Pretraining with Active Forgetting paper page: https://
huggingface.co/papers/2307.01
163
… Pretrained language models (PLMs) are today the primary model for natural language processing. Despite their impressive downstream performance, it can be difficult to apply
@_akhaliq
-
Improving Language Plasticity via Pretraining with Active Forgetting
By
–
-
ChatGPT Midjourney Photoshop Create Novel Emotions
By
–
ChatGPT + Midjourney + Photoshop, invent new emotions by @jphilipp 1.
-
Statler: State-Maintaining Language Models for Embodied Reasoning
By
–
Statler: State-Maintaining Language Models for Embodied Reasoning paper page: https://
huggingface.co/papers/2306.17
840
… Large language models (LLMs) provide a promising tool that enable robots to perform complex robot reasoning tasks. However, the limited context window of contemporary LLMs -
ReMaX: Relaxing Mask Transformers for Efficient Panoptic Segmentation
By
–
ReMaX: Relaxing for Better Training on Efficient Panoptic Segmentation paper page: https://
huggingface.co/papers/2306.17
319
… paper presents a new mechanism to facilitate the training of mask transformers for efficient panoptic segmentation, democratizing its deployment. We observe that due to -
Hardwiring ViT Patch Selectivity into CNNs using Patch Mixing
By
–
Hardwiring ViT Patch Selectivity into CNNs using Patch Mixing paper page: https://
huggingface.co/papers/2306.17
848
… Vision transformers (ViTs) have significantly changed the computer vision landscape and have periodically exhibited superior performance in vision tasks compared to convolutional -
Shaped Transformer: Attention Models in Infinite Depth-Width Limit
By
–
The Shaped Transformer: Attention Models in the Infinite Depth-and-Width Limit paper page: https://
huggingface.co/papers/2306.17
759
… In deep learning theory, the covariance matrix of the representations serves as a proxy to examine the network's trainability. Motivated by the success of -
Magic123: Generate High-Quality 3D Objects from Single Images
By
–
Magic123: One Image to High-Quality 3D Object Generation Using Both 2D and 3D Diffusion Priors
— AK (@_akhaliq) 3 juillet 2023
paper page: https://t.co/h869cnIX32
present Magic123, a two-stage coarse-to-fine approach for high-quality, textured 3D meshes generation from a single unposed image in the wild using… pic.twitter.com/nnhlRzqDMnMagic123: One Image to High-Quality 3D Object Generation Using Both 2D and 3D Diffusion Priors paper page: https://
huggingface.co/papers/2306.17
843
… present Magic123, a two-stage coarse-to-fine approach for high-quality, textured 3D meshes generation from a single unposed image in the wild using -
Classifier-Free Guidance Improves Prompt Adherence in Image Generation
By
–
Stay on topic with Classifier-Free Guidance paper page: https://
huggingface.co/papers/2306.17
806
… Classifier-Free Guidance (CFG) has recently emerged in text-to-image generation as a lightweight technique to encourage prompt-adherence in generations. In this work, we demonstrate that CFG can be -
Exploitability of Instruction Tuning in Large Language Models
By
–
On the Exploitability of Instruction Tuning paper page: https://
huggingface.co/papers/2306.17
194
… Instruction tuning is an effective technique to align large language models (LLMs) with human intents. In this work, we investigate how an adversary can exploit instruction tuning by injecting -
ChatGPT for Robotics: Design Principles and Model Abilities
By
–
ChatGPT for Robotics: Design Principles and Model Abilities paper page: https://
huggingface.co/papers/2306.17
582
… paper presents an experimental study regarding the use of OpenAI's ChatGPT for robotics applications. We outline a strategy that combines design principles for prompt engineering and