John Schulman - Reinforcement Learning from Human Feedback: Progress and Challenges

These chapters are auto-generated

EECS Colloquium Wednesday, April 19, 2023 Banatao Auditorium 5-6p

Introduction

0:00

Overview

2:06

Hallucination

3:35

Conceptual Model

8:47

Transcript

Follow along using the transcript.

UC Berkeley EECS

10.1K subscribers

Live chat replay is not available for this video.

UC Open 2025 vLLM: A success story of UC and industry collaboration in Open Source AI

Yelena Martynovskaya

•

8 views 8 days ago

Deep Learning Basics: Introduction and Overview

Lex Fridman

•

2.4M views 6 years ago

Putting Third Party Data to Work with Your Telco Churn Models 1

Mobilewalla

•

1 view 20 hours ago

New

Reinforcement Learning from Human Feedback (RLHF) Explained

IBM Technology

•

44K views 9 months ago

MIT EI seminar, Hyung Won Chung from OpenAI. "Don't teach. Incentivize."

Hyung Won Chung

•

38K views 8 months ago

The FASTEST introduction to Reinforcement Learning on the internet

Gonkee

•

59K views 5 months ago

John Schulman - Reinforcement Learning from Human Feedback: Progress and Challenges

Chapters View all

Introduction

Introduction

Introduction

Overview

Overview

Overview

Hallucination

Hallucination

Hallucination

Conceptual Model

Conceptual Model

Conceptual Model

Behavior Cloning

Behavior Cloning

Behavior Cloning

Does the model know

Does the model know

Does the model know

Uncertainty

Uncertainty

Uncertainty

When should you hedge

When should you hedge

When should you hedge

UC Berkeley EECS

John Schulman - Reinforcement Learning from Human Feedback: Progress and Challenges

Comments 28

Chapters

Introduction

Introduction

Introduction

Overview

Overview

Overview

Hallucination

Hallucination

Hallucination

Conceptual Model

Conceptual Model

Conceptual Model

Behavior Cloning

Behavior Cloning

Behavior Cloning

Does the model know

Does the model know

Does the model know

Uncertainty

Uncertainty

Uncertainty

When should you hedge

When should you hedge

When should you hedge

Long form answers

Long form answers

Long form answers

Improving factuality

Improving factuality

Improving factuality

Challenges

Challenges

Challenges

Retrieval Citing Sources

Retrieval Citing Sources

Retrieval Citing Sources

RL Environment

RL Environment

RL Environment

RL Task

RL Task

RL Task

RL Pipeline

RL Pipeline

RL Pipeline

Browsing

Browsing

Browsing

Dagger

Dagger

Chapters

Chapters