Reinforced Self-Training (ReST) for Language Modeling (Paper Explained)
#ai #rlhf #llm
ReST uses a bootsrap-like method to produce its own extended dataset and trains on ever higher-quality subsets of it to improve its own reward. The method allows for re-using the same generated data multiple times and thus has an efficiency advantage with respect to Online RL techniques like PPO.
Paper: https://arxiv.org/abs/2308.08998
Abstract:
Reinforcement learning from human feedback (RLHF) can improve the quality of large language model's (LLM) outputs by aligning them with human preferences. We propose a simple algorithm for aligning LLMs with human preferences inspired by growing batch reinforcement learning (RL), which we call Reinforced Self-Training (ReST). Given an initial LLM policy, ReST produces a dataset by generating samples from the policy, which are then used to improve the LLM policy using offline RL algorithms. ReST is more efficient than typical online RLHF methods because the training dataset is produced offline, which allows data reuse. While ReST is a general approach applicable to all generative learning settings, we focus on its application to machine translation. Our results show that ReST can substantially improve translation quality, as measured by automated metrics and human evaluation on machine translation benchmarks in a compute and sample-efficient manner.
Authors: Caglar Gulcehre, Tom Le Paine, Srivatsan Srinivasan, Ksenia Konyushkova, Lotte Weerts, Abhishek Sharma, Aditya Siddhant, Alex Ahern, Miaosen Wang, Chenjie Gu, Wolfgang Macherey, Arnaud Doucet, Orhan Firat, Nando de Freitas
Links:
Homepage: https://ykilcher.com
Merch: https://ykilcher.com/merch
YouTube: https://www.youtube.com/c/yannickilcher
Twitter: https://twitter.com/ykilcher
Discord: https://ykilcher.com/discord
LinkedIn: https://www.linkedin.com/in/ykilcher
If you want to support me, the best thing to do is to share out the content :)
If you want to support me financially (completely optional and voluntary, but a lot of people have asked for this):
SubscribeStar: https://www.subscribestar.com/yannickilcher
Patreon: https://www.patreon.com/yannickilcher
Bitcoin (BTC): bc1q49lsw3q325tr58ygf8sudx2dqfguclvngvy2cq
Ethereum (ETH): 0x7ad3513E3B8f66799f507Aa7874b1B0eBC7F85e2
Litecoin (LTC): LQW2TRyKYetVC8WjFkhpPhtpbDM4Vw7r9m
Monero (XMR): 4ACL8AGrEo5hAir8A9CeVrW8pEauWvnp1WnSDZxW7tziCDLhZAGsgzhRQABDnFy8yuM9fWJDviJPHKRjV4FWt19CJZN9D4n
Видео Reinforced Self-Training (ReST) for Language Modeling (Paper Explained) канала Yannic Kilcher
ReST uses a bootsrap-like method to produce its own extended dataset and trains on ever higher-quality subsets of it to improve its own reward. The method allows for re-using the same generated data multiple times and thus has an efficiency advantage with respect to Online RL techniques like PPO.
Paper: https://arxiv.org/abs/2308.08998
Abstract:
Reinforcement learning from human feedback (RLHF) can improve the quality of large language model's (LLM) outputs by aligning them with human preferences. We propose a simple algorithm for aligning LLMs with human preferences inspired by growing batch reinforcement learning (RL), which we call Reinforced Self-Training (ReST). Given an initial LLM policy, ReST produces a dataset by generating samples from the policy, which are then used to improve the LLM policy using offline RL algorithms. ReST is more efficient than typical online RLHF methods because the training dataset is produced offline, which allows data reuse. While ReST is a general approach applicable to all generative learning settings, we focus on its application to machine translation. Our results show that ReST can substantially improve translation quality, as measured by automated metrics and human evaluation on machine translation benchmarks in a compute and sample-efficient manner.
Authors: Caglar Gulcehre, Tom Le Paine, Srivatsan Srinivasan, Ksenia Konyushkova, Lotte Weerts, Abhishek Sharma, Aditya Siddhant, Alex Ahern, Miaosen Wang, Chenjie Gu, Wolfgang Macherey, Arnaud Doucet, Orhan Firat, Nando de Freitas
Links:
Homepage: https://ykilcher.com
Merch: https://ykilcher.com/merch
YouTube: https://www.youtube.com/c/yannickilcher
Twitter: https://twitter.com/ykilcher
Discord: https://ykilcher.com/discord
LinkedIn: https://www.linkedin.com/in/ykilcher
If you want to support me, the best thing to do is to share out the content :)
If you want to support me financially (completely optional and voluntary, but a lot of people have asked for this):
SubscribeStar: https://www.subscribestar.com/yannickilcher
Patreon: https://www.patreon.com/yannickilcher
Bitcoin (BTC): bc1q49lsw3q325tr58ygf8sudx2dqfguclvngvy2cq
Ethereum (ETH): 0x7ad3513E3B8f66799f507Aa7874b1B0eBC7F85e2
Litecoin (LTC): LQW2TRyKYetVC8WjFkhpPhtpbDM4Vw7r9m
Monero (XMR): 4ACL8AGrEo5hAir8A9CeVrW8pEauWvnp1WnSDZxW7tziCDLhZAGsgzhRQABDnFy8yuM9fWJDviJPHKRjV4FWt19CJZN9D4n
Видео Reinforced Self-Training (ReST) for Language Modeling (Paper Explained) канала Yannic Kilcher
Показать
Комментарии отсутствуют
Информация о видео
Другие видео канала
OpenAssistant is Completed[ML News] GPT-4 solves MIT Exam with 100% ACCURACY | OpenLLaMA 13B releasedTree-Ring Watermarks: Fingerprints for Diffusion Images that are Invisible and Robust (Explained)RWKV: Reinventing RNNs for the Transformer Era (Paper Explained)Tree of Thoughts: Deliberate Problem Solving with Large Language Models (Full Paper Review)OpenAI suggests AI licenses (US Senate hearing on AI regulation w/ Sam Altman)[ML News] Geoff Hinton leaves Google | Google has NO MOAT | OpenAI down half a billionScaling Transformer to 1M tokens and beyond with RMT (Paper Explained)AI Alignment Livestream (aka OpenAssistant "Just Chatting")OpenAssistant First Models are here! (Open-Source ChatGPT)The biggest week in AI (GPT-4, Office Copilot, Google PaLM, Anthropic Claude & more)GPT-4 is here! What we know so far (Full Analysis)This ChatGPT Skill will earn you $10B (also, AI reads your mind!) | ML NewsLLaMA: Open and Efficient Foundation Language Models (Paper Explained)Open Assistant Inference Backend Development (Hands-On Coding)OpenAssistant - ChatGPT's Open Alternative (We need your help!)Open Assistant Live Coding (Open-Source ChatGPT Replication)AI Essay Competition (lab42)Open Assistant Live Coding (Open-Source ChatGPT Replication)ChatGPT: This AI has a JAILBREAK?! (Unbelievable AI Progress)