Cerebras-GPT – Deploying the first open-source compute-optimal models – https://
hubs.li/Q01Z9tgQ0
RESEARCH
-
Cerebras-GPT: First Open-Source Compute-Optimal Models
By
–
-
Sparse-IFT: Larger Sparse Models with Higher Accuracy
By
–
Sparse-IFT – Creating larger sparse models with higher accuracy at the same compute – https://
hubs.li/Q01Z9vCZ0 -
VSL: Variable Sequence Lengths for Cost-Effective Long Context
By
–
VSL – Utilizing variable sequence lengths to achieve longer sequences at lower cost – https://
hubs.li/Q01Z9jxc0 -
SparseGPT: Sparsifying LLMs for efficient inference
By
–
SparseGPT – Sparsifying LLMs for efficient inference – https://
hubs.li/Q01Z9ysx0 -
Sparse models match dense accuracy with fewer flops
By
–
SPDF – Matching the downstream accuracy of a dense model with a sparse model using fewer flops – https://
hubs.li/Q01Z9v0G0 -
Visual Foundation Models: Architectures, Training, and Prompting Survey
By
–
Foundational Models Defining a New Era in Vision: A Survey and Outlook A comprehensive survey of visual foundation models. Covers emerging architectures(for vision+other modalities), training objectives, pre-training datasets, fine-tuning mechanisms, and the common prompting
-
LLM Text Adversarial Problems Easier Than Vision Recognition
By
–
I do agree that trying to solve this within LLM will be hard. But the problem seems easier to solve in practice than vision, unless people come up with a way to make an adversarial string that reads like natural language
-
Adversarial Examples in Vision vs Language Models Vulnerability
By
–
Adversarial examples have been hard to solve in vision mainly because it was imperceptible to humans and added adv noise is hard to remove. Looking at the demo, it seems relatively easy to build a filter to remove adversarial suffix before feeding the query into LLM…
-
NVIDIA AI Algorithms Team Coffee Chat Event
By
–
Our AI Algorithms team at @nvidia is having a coffee chat this evening. DM us for details. @yukez @DrJimFan @ChaoweiX @Azizzadenesheli @JeanKossaifi
-
Think Week Retreats for AI Leaders and Innovators
By
–
Bill Gates used to go offline to do a Think Week twice a year. All he'd do is read, think, and plan. Would you do one if we organized it through @every
? 3-7 days in a new location with like-minded people—lots of time to let your mind roam, and share ideas.