Your Daily AI Research tl;dr — 2022–10–17 🧠

Introducing LAION-5B: a dataset of 5,85 billion CLIP-filtered image-text pairs, RL for story generation without human annotations and more…

Welcome to your official daily AI research tl;dr (often with code and news) for AI professionals where I share the most exciting papers I find daily, along with a one-liner summary to help you quickly determine if the article (and code) is worth investigating.

1️⃣ Robust Preference Learning for Storytelling via Contrastive Reinforcement Learning

Reinforcement learning for story generation but using no human annotation data! “We explore the limitations of CLIP-like story critic models aka CARP and show various improvements that can be made in order to improve robustness, an essential factor in applying online reinforcement learning.”

Link to the paper: https://arxiv.org/pdf/2210.07792.pdf

2️⃣ MTEB: Massive Text Embedding Benchmark

“We introduce the Massive Text Embedding Benchmark (MTEB) that spans 8 embedding tasks covering a total of 56 datasets and 112 languages.”