With the rapid development of language technology, it’s important that as many languages as possible benefit from these technologies, so we’re sharing XTREME-UP, a benchmark for evaluating multilingual models. https://
goo.gle/xtreme-up-paper https://
github.com/google-researc
h/xtreme-up
…
Read ↓ (1/3)
@googleai
-
Google Releases XTREME-UP Multilingual Model Evaluation Benchmark
By
–
-
Differential Privacy in Machine Learning: Techniques and Challenges
By
–
Today, we discuss the current state of differentially private ML (DP-ML) research with an overview of common techniques for obtaining DP-ML models, engineering challenges, mitigation techniques and current open questions. Learn more ↓
-
Vision Transformer Efficient Video Backbone Using Sparse Tubes
By
–
Learn how we turned a Vision Transformer image encoder into an efficient video backbone using sparse video tubes (3D grid-based cuboids with learnable visual representations of video samples), reducing compute needs and achieving competitive results → https://
goo.gle/435CTYQ -
Google PAIR Team Advances Human-AI Interaction Research and Tools
By
–
Today we describe the many efforts of the People + AI Research (PAIR) team within Google Research, which conducts foundational work on human-AI interaction to generate educational materials, tools and software that change the way people engage with AI. https://t.co/9I2nGjYMwG pic.twitter.com/GADbB2kPWi
— Google AI (@GoogleAI) 18 mai 2023Today we describe the many efforts of the People + AI Research (PAIR) team within Google Research, which conducts foundational work on human-AI interaction to generate educational materials, tools and software that change the way people engage with AI. https://
goo.gle/41QpjYr -
Dynamic Planning Advances for Human-Assistant Conversations
By
–
Read all about our recent advances in dynamic planning for human-to-assistant conversations, in which we enable an assistant to plan a multi-turn conversation toward a goal and adapt that plan in real-time via a #ReinforcementLearning-based approach ↓
-
In-Context Learning Scaling in Large Language Models
By
–
During in-context learning (ICL), models are prompted with a few examples of input-label pairs before performing a task on an unseen example. Read how larger language models do in-context learning differently & how this can change with their scale → https://
goo.gle/3Mwflag -
Google releases Monk Skin Tone dataset for inclusive AI
By
–
Reliable & accurate skin tone annotations are crucial for computer vision, so we’re building more inclusive skin tone scales and datasets. Today, we’re releasing the Monk Skin Tone Examples dataset so others can ensure their technology sees everyone too. https://
goo.gle/3ObepsM -
F-VLM: Open-Vocabulary Detection with Frozen Vision Language Models
By
–
F-VLM, a simple and scalable open-vocabulary detection method that is built upon frozen vision and language models, reduces the training complexity for open-vocabulary detectors and expands detection to novel objects. Learn more and check out the code → https://
goo.gle/3O6Ih9Y -
Prompting Techniques Adapt LLMs Mobile User Interfaces
By
–
Read how our suite of prompting techniques, combined with a novel algorithm, can adapt LLMs to mobile user interfaces (UIs), enabling various conversational interactions with mobile UIs while achieving competitive results and saving developers time. https://t.co/9ABtjtylnS pic.twitter.com/crZ01VyRip
— Google AI (@GoogleAI) 12 mai 2023Read how our suite of prompting techniques, combined with a novel algorithm, can adapt LLMs to mobile user interfaces (UIs), enabling various conversational interactions with mobile UIs while achieving competitive results and saving developers time. https://
goo.gle/42w91Vs -
Google Med-PaLM: Advanced Medical AI Language Model
By
–
Want to learn more about Med-PaLM? Read about it and watch the video at https://
sites.research.google/med-palm/