Papers I Read Notes and Summaries

Toolformer - Language Models Can Teach Themselves to Use Tools

Introduction

  • The paper presents Toolformer, a language model that uses simple APIs...


Synthesized Policies for Transfer and Adaptation across Tasks and Environments

Introduction

  • The paper studies transfer learning in RL, focusing on simultaneous transfer...


Deep Neural Networks for YouTube Recommendations

Introduction

  • The paper describes YouTube’s deep learning-based recommendation system.

  • Continue reading -->


The Tail at Scale

Introduction

  • The paper presents some causes for (temporary) high-latency episodes in large-scale...


Practical Lessons from Predicting Clicks on Ads at Facebook

Introduction

  • The paper describes several design choices for developing a model for...


Ad Click Prediction - a View from the Trenches

Introduction

  • The paper presents case studies from the experience of deploying an...


Anatomy of Catastrophic Forgetting - Hidden Representations and Task Semantics

Introduction

  • The paper studies the effect of catastrophic forgetting on representations in...


When Do Curricula Work?

Introduction

  • The paper systematically investigates when does curriculum learning help.

  • ...

Continual learning with hypernetworks

Introduction

  • The paper proposes the use of task-conditioned HyperNetworks for lifelong...


Zero-shot Learning by Generating Task-specific Adapters

Introduction

  • The paper introduces HYPTER - a framework for zero-shot learning (ZSL)...