Intrinsically-Motivated Humans and Agents in Open-World Exploration

What drives exploration? Understanding intrinsic motivation is a long-standing challenge in both cognitive science and artificial intelligence; numerous objectives have been proposed and used to train agents, yet there remains a gap between human and agent exploration. We directly compare adults, children, and AI agents in a complex open-ended environment, Crafter, and study how common intrinsic objectives: Entropy, Information Gain, and Empowerment, relate to their behavior. We find that only Entropy and Empowerment are consistently positively correlated with human exploration progress, indicating that these objectives may better inform intrinsic reward design for agents. Furthermore, across agents and humans we observe that Entropy initially increases rapidly, then plateaus, while Empowerment increases continuously, suggesting that state diversity may provide more signal in early exploration, while advanced exploration should prioritize control. Finally, we find preliminary evidence that private speech utterances, and particularly goal verbalizations, may aid exploration in children.

Imagen 3

We introduce Imagen 3, a latent diffusion model that generates high quality images from text prompts. We describe our quality and responsibility evaluations. Imagen 3 is preferred over other state-of-the-art (SOTA) models at the time of evaluation. In addition, we discuss issues around safety and representation, as well as methods we used to minimize the potential harm of our models.

Semi-Supervised One-Shot Imitation Learning

One-shot Imitation Learning~(OSIL) aims to imbue AI agents with the ability to learn a new task from a single demonstration. To supervise the learning, OSIL typically requires a prohibitively large number of paired expert demonstrations -- i.e. trajectories corresponding to different variations of the same semantic task. To overcome this limitation, we introduce the semi-supervised OSIL problem setting, where the learning agent is presented with a large dataset of trajectories with no task labels (i.e. an unpaired dataset), along with a small dataset of multiple demonstrations per semantic task (i.e. a paired dataset). This presents a more realistic and practical embodiment of few-shot learning and requires the agent to effectively leverage weak supervision from a large dataset of trajectories. Subsequently, we develop an algorithm specifically applicable to this semi-supervised OSIL setting. Our approach first learns an embedding space where different tasks cluster uniquely. We utilize this embedding space and the clustering it supports to self-generate pairings between trajectories in the large unpaired dataset. Through empirical results on simulated control tasks, we demonstrate that OSIL models trained on such self-generated pairings are competitive with OSIL models trained with ground-truth labels, presenting a major advancement in the label-efficiency of OSIL.

Teaching Large Language Models to Reason with Reinforcement Learning

Reinforcement Learning from Human Feedback (\textbf{RLHF}) has emerged as a dominant approach for aligning LLM outputs with human preferences. Inspired by the success of RLHF, we study the performance of multiple algorithms that learn from feedback (Expert Iteration, Proximal Policy Optimization (\textbf{PPO}), Return-Conditioned RL) on improving LLM reasoning capabilities. We investigate both sparse and dense rewards provided to the LLM both heuristically and via a learned reward model. We additionally start from multiple model sizes and initializations both with and without supervised fine-tuning (\textbf{SFT}) data. Overall, we find all algorithms perform comparably, with Expert Iteration performing best in most cases. Surprisingly, we find the sample complexity of Expert Iteration is similar to that of PPO, requiring at most on the order of $10^6$ samples to converge from a pretrained checkpoint. We investigate why this is the case, concluding that during RL training models fail to explore significantly beyond solutions already produced by SFT models. Additionally, we discuss a trade off between maj@1 and pass@96 metric performance during SFT training and how conversely RL training improves both simultaneously. We then conclude by discussing the implications of our findings for RLHF and the future role of RL in LLM fine-tuning.

Learning to Model the World with Language

To interact with humans in the world, agents need to understand the diverse types of language that people use, relate them to the visual world, and act based on them. While current agents learn to execute simple language instructions from task rewards, we aim to build agents that leverage diverse language that conveys general knowledge, describes the state of the world, provides interactive feedback, and more. Our key idea is that language helps agents predict the future: what will be observed, how the world will behave, and which situations will be rewarded. This perspective unifies language understanding with future prediction as a powerful self-supervised learning objective. We present Dynalang, an agent that learns a multimodal world model that predicts future text and image representations and learns to act from imagined model rollouts. Unlike traditional agents that use language only to predict actions, Dynalang acquires rich language understanding by using past language also to predict future language, video, and rewards. In addition to learning from online interaction in an environment, Dynalang can be pretrained on datasets of text, video, or both without actions or rewards. From using language hints in grid worlds to navigating photorealistic scans of homes, Dynalang utilizes diverse types of language to improve task performance, including environment descriptions, game rules, and instructions.

DPOK: Reinforcement Learning for Fine-tuning Text-to-Image Diffusion Models

Learning from human feedback has been shown to improve text-to-image models. These techniques first learn a reward function that captures what humans care about in the task and then improve the models based on the learned reward function. Even though relatively simple approaches (e.g., rejection sampling based on reward scores) have been investigated, fine-tuning text-to-image models with the reward function remains challenging. In this work, we propose using online reinforcement learning (RL) to fine-tune text-to-image models. We focus on diffusion models, defining the fine-tuning task as an RL problem, and updating the pre-trained text-to-image diffusion models using policy gradient to maximize the feedback-trained reward. Our approach, coined DPOK, integrates policy optimization with KL regularization. We conduct an analysis of KL regularization for both RL fine-tuning and supervised fine-tuning. In our experiments, we show that DPOK is generally superior to supervised fine-tuning with respect to both image-text alignment and image quality.

Vision-Language Models as Success Detectors

Detecting successful behaviour is crucial for training intelligent agents. As such, generalisable reward models are a prerequisite for agents that can learn to generalise their behaviour. In this work we focus on developing robust success detectors that leverage large, pretrained vision-language models (Flamingo, Alayrac et al. (2022)) and human reward annotations. Concretely, we treat success detection as a visual question answering (VQA) problem, denoted SuccessVQA. We study success detection across three vastly different domains: (i) interactive language-conditioned agents in a simulated household, (ii) real world robotic manipulation, and (iii) "in-the-wild" human egocentric videos. We investigate the generalisation properties of a Flamingo-based success detection model across unseen language and visual changes in the first two domains, and find that the proposed method is able to outperform bespoke reward models in out-of-distribution test scenarios with either variation. In the last domain of "in-the-wild" human videos, we show that success detection on unseen real videos presents an even more challenging generalisation task warranting future work. We hope our initial results encourage further work in real world success detection and reward modelling.

Aligning Text-to-Image Models using Human Feedback

Deep generative models have shown impressive results in text-to-image synthesis. However, current text-to-image models often generate images that are inadequately aligned with text prompts. We propose a fine-tuning method for aligning such models using human feedback, comprising three stages. First, we collect human feedback assessing model output alignment from a set of diverse text prompts. We then use the human-labeled image-text dataset to train a reward function that predicts human feedback. Lastly, the text-to-image model is fine-tuned by maximizing reward-weighted likelihood to improve image-text alignment. Our method generates objects with specified colors, counts and backgrounds more accurately than the pre-trained model. We also analyze several design choices and find that careful investigations on such design choices are important in balancing the alignment-fidelity tradeoffs. Our results demonstrate the potential for learning from human feedback to significantly improve text-to-image models.

Guiding Pretraining in Reinforcement Learning with Large Language Models

Reinforcement learning algorithms typically struggle in the absence of a dense, well-shaped reward function. Intrinsically motivated exploration methods address this limitation by rewarding agents for visiting novel states or transitions, but these methods offer limited benefits in large environments where most discovered novelty is irrelevant for downstream tasks. We describe a method that uses background knowledge from text corpora to shape exploration. This method, called ELLM (Exploring with LLMs) rewards an agent for achieving goals suggested by a language model prompted with a description of the agent's current state. By leveraging large-scale language model pretraining, ELLM guides agents toward human-meaningful and plausibly useful behaviors without requiring a human in the loop. We evaluate ELLM in the Crafter game environment and the Housekeep robotic simulator, showing that ELLM-trained agents have better coverage of common-sense behaviors during pretraining and usually match or improve performance on a range of downstream tasks.

It Takes Four to Tango: Multiagent Selfplay for Automatic Curriculum Generation

We are interested in training general-purpose reinforcement learning agents that can solve a wide variety of goals. Training such agents efficiently requires automatic generation of a goal curriculum. This is challenging as it requires (a) exploring goals of increasing difficulty, while ensuring that the agent (b) is exposed to a diverse set of goals in a sample efficient manner and (c) does not catastrophically forget previously solved goals. We propose Curriculum Self Play (CuSP), an automated goal generation framework that seeks to satisfy these desiderata by virtue of a multi-player game with four agents. We extend the asymmetric curricula learning in PAIRED (Dennis et al., 2020) to a symmetrized game that carefully balances cooperation and competition between two off-policy student learners and two regret-maximizing teachers. CuSP additionally introduces entropic goal coverage and accounts for the non-stationary nature of the students, allowing us to automatically induce a curriculum that balances progressive exploration with anti-catastrophic exploitation. We demonstrate that our method succeeds at generating an effective curricula of goals for a range of control tasks, outperforming other methods at zero-shot test-time generalization to novel out-of-distribution goals.