EPISODE · May 15, 2024 · 14 MIN
Google I/O 📣 // MS MARCO Web Search Dataset 🔍 // Vision-Language Models 🌅
from GPT Reviews · host Earkind
Google I/O 2024 announcements, including new AI tools like Firebase Genkit, LearnLM, and Veo, as well as Gemini, an AI replacement for Google Assistant. The introduction of the MS MARCO Web Search dataset, which provides a retrieval benchmark with three web retrieval challenge tasks and millions of real-clicked query-document pairs for training and evaluating retrieval models. The "What matters when building vision-language models?" paper, which identifies critical decisions in the design of vision-language models and presents Idefics2, an efficient foundational VLM of 8 billion parameters that achieves state-of-the-art performance within its size category. The "RLHF Workflow: From Reward Modeling to Online RLHF" paper, which presents a workflow for Online Iterative Reinforcement Learning from Human Feedback (RLHF) in an online setting and achieves impressive performance on LLM chatbot benchmarks and academic benchmarks. Contact: [email protected] Timestamps: 00:34 Introduction 01:34 Google I/O 2024: Here’s everything Google just announced 03:26 Ilya Sutskever leaves OpenAI 04:57 GPT-4o’s Memory Breakthrough! 06:00 Fake sponsor 07:49 MS MARCO Web Search: a Large-scale Information-rich Web Dataset with Millions of Real Click Labels 09:33 What matters when building vision-language models? 10:54 RLHF Workflow: From Reward Modeling to Online RLHF 13:00 Outro
Embed this episode
NOW PLAYING
Google I/O 📣 // MS MARCO Web Search Dataset 🔍 // Vision-Language Models 🌅
No transcript for this episode yet
Similar Episodes
No similar episodes found.
Similar Podcasts
No similar podcasts found.