Michael Noukhovitch

Asynchronous RLHF: Faster and More Efficient Off-Policy RL for Language Models

Shengyi Huang

The dominant paradigm for RLHF is online and on-policy RL: synchronously generating from the large language model (LLM) policy, labelling wi… (see more)th a reward model, and learning using feedback on the LLM's own outputs. While performant, this paradigm is computationally inefficient. Inspired by classical deep RL literature, we propose separating generation and learning in RLHF. This enables asynchronous generation of new samples while simultaneously training on old samples, leading to faster training and more compute-optimal scaling. However, asynchronous training relies on an underexplored regime, online but off-policy RLHF: learning on samples from previous iterations of our model which give a worse training signal. We tackle the fundamental challenge in this regime: how much off-policyness can we tolerate for asynchronous training to speed up learning but maintain performance? Among several RLHF algorithms we test, online DPO is found to be most robust to off-policy data, and robustness increases with the scale of the policy model. We study further compute optimizations for asynchronous RLHF but find that they come at a performance cost, giving rise to a trade-off. We verify the scalability of asynchronous RLHF by training a general-purpose chatbot from LLaMA 3.1 8B on an instruction-following task ~40% faster than a synchronous run while matching final performance. Finally, we extend our results to math and reasoning to demonstrate asynchronous RL can finetune Rho 1B on GSM8k ~70% faster while matching synchronous accuracy.

2025-01-21

ICLR.cc/2025/Conference (poster)

In-Context Learning, Can It Break Safety?

David Dobre

2024-06-27

ICML.cc/2024/Workshop/NextGenAISafety (poster)

Language Model Alignment with Elastic Reset

Samuel Lavoie

Florian Strub

Finetuning language models with reinforcement learning (RL), e.g. from human feedback (HF), is a prominent method for alignment. But optimiz… (see more)ing against a reward model can improve on reward while degrading performance in other areas, a phenomenon known as reward hacking, alignment tax, or language drift. First, we argue that commonly-used test metrics are insufficient and instead measure how different algorithms tradeoff between reward and drift. The standard method modified the reward with a Kullback-Lieber (KL) penalty between the online and initial model. We propose Elastic Reset, a new algorithm that achieves higher reward with less drift without explicitly modifying the training objective. We periodically reset the online model to an exponentially moving average (EMA) of itself, then reset the EMA model to the initial model. Through the use of an EMA, our model recovers quickly after resets and achieves higher reward with less drift in the same number of steps. We demonstrate that fine-tuning language models with Elastic Reset leads to state-of-the-art performance on a small scale pivot-translation benchmark, outperforms all baselines in a medium-scale RLHF-like IMDB mock sentiment task and leads to a more performant and more aligned technical QA chatbot with LLaMA-7B. Code available at github.com/mnoukhov/elastic-reset.

2023-09-20

NeurIPS.cc/2023/Conference (poster)

Simplicial Embeddings in Self-Supervised Learning and Downstream Classification

Samuel Lavoie

Simplicial Embeddings (SEM) are representations learned through self-supervised learning (SSL), wherein a representation is projected into …

2023-01-31

ICLR.cc/2023/Conference (notable)

Pretraining Reward-Free Representations for Data-Efficient Reinforcement Learning

Philip Bachman

2021-03-08

International Conference on Learning Representations (unknown)

DATA-EFFICIENT REINFORCEMENT LEARNING

R Devon Hjelm

Philip Bachman

Data efficiency poses a major challenge for deep reinforcement learning. We approach this issue from the perspective of self-supervised repr… (see more)esentation learning, leveraging reward-free exploratory data to pretrain encoder networks. We employ a novel combination of latent dynamics modelling and goal-reaching objectives, which exploit the inherent structure of data in reinforcement learning. We demonstrate that our method scales well with network capacity and pretraining data. When evaluated on the Atari 100k data-efficiency benchmark, our approach significantly outperforms previous methods combining unsupervised pretraining with task-specific finetuning, and approaches human-level performance.

2020-12-31

(published)

www.semanticscholar.org

Emergent Communication under Competition

Travis LaCroix

Angeliki Lazaridou

2020-12-31

AAMAS (published)

arxiv.org

Pretraining Representations for Data-Efficient Reinforcement Learning

Philip Bachman

Data efficiency is a key challenge for deep reinforcement learning. We address this problem by using unlabeled data to pretrain an encoder w… (see more)hich is then finetuned on a small amount of task-specific data. To encourage learning representations which capture diverse aspects of the underlying MDP, we employ a combination of latent dynamics modelling and unsupervised goal-conditioned RL. When limited to 100k steps of interaction on Atari games (equivalent to two hours of human experience), our approach significantly surpasses prior work combining offline representation pretraining with task-specific finetuning, and compares favourably with other pretraining methods that require orders of magnitude more data. Our approach shows particular promise when combined with larger models as well as more diverse, task-aligned observational data -- approaching human-level performance and data-efficiency on Atari in our best setting. We provide code associated with this work at https://github.com/mila-iqia/SGI.

2020-12-31

Advances in Neural Information Processing Systems 34 (NeurIPS 2021) (published)

Selfish Emergent Communication

Travis LaCroix

2019-09-24

(published)

Systematic Generalization: What Is Required and Can It Be Learned?

Dzmitry Bahdanau*

Shikhar Murty*

2018-12-31

ICLR.cc/2019/Conference (poster)

Commonsense mining as knowledge base completion? A study on the impact of novelty

Stanisław Jastrzębski

Dzmitry Bahdanau

Seyedarian Hosseini

Yoshua Bengio

Jackie Chi Kit Cheung

Commonsense knowledge bases such as ConceptNet represent knowledge in the form of relational triples. Inspired by the recent work by Li et a… (see more)l., we analyse if knowledge base completion models can be used to mine commonsense knowledge from raw text. We propose novelty of predicted triples with respect to the training set as an important factor in interpreting results. We critically analyse the difficulty of mining novel commonsense knowledge, and show that a simple baseline method outperforms the previous state of the art on predicting more novel.

2018-05-31

Proceedings of the Workshop on Generalization in the Age of Deep Learning (published)