Large language models (LLMs) may not equitably represent diverse global perspectives on societal issues. In this paper, we develop a quantitative framework to evaluate whose opinions model-generated responses are more similar to. We first build a dataset, GlobalOpinionQA, comprised of questions and answers from cross-national surveys designed to capture diverse opinions on global issues across different countries. Next, we define a metric that quantifies the similarity between LLM-generated survey responses and human responses, conditioned on country. With our framework, we run three experiments on an LLM trained to be helpful, honest, and harmless with Constitutional AI. By default, LLM responses tend to be more similar to the opinions of certain populations, such as those from the USA, and some European and South American countries, highlighting the potential for biases. When we prompt the model to consider a particular country's perspective, responses shift to be more similar to the opinions of the prompted populations, but can reflect harmful cultural stereotypes. When we translate GlobalOpinionQA questions to a target language, the model's responses do not necessarily become the most similar to the opinions of speakers of those languages. We release our dataset for others to use and build on. Our data is at https://huggingface.co/datasets/Anthropic/llm_global_opinions. We also provide an interactive visualization at https://llmglobalvalues.anthropic.com.

Fully cooperative, partially observable multi-agent problems are ubiquitous in the real world. In this paper, we focus on a specific subclass of coordination problems in which humans are able to discover self-explaining deviations (SEDs). SEDs are actions that deviate from the common understanding of what reasonable behavior would be in normal circumstances. They are taken with the intention of causing another agent or other agents to realize, using theory of mind, that the circumstance must be abnormal. We first motivate SED with a real world example and formalize its definition. Next, we introduce a novel algorithm, improvement maximizing self-explaining deviations (IMPROVISED), to perform SEDs. Lastly, we evaluate IMPROVISED both in an illustrative toy setting and the popular benchmark setting Hanabi, where it is the first method to produce so called finesse plays, which are regarded as one of the more iconic examples of human theory of mind.

We consider the task of building strong but human-like policies in multi-agent decision-making problems, given examples of human behavior. Imitation learning is effective at predicting human actions but may not match the strength of expert humans, while self-play learning and search techniques (e.g. AlphaZero) lead to strong performance but may produce policies that are difficult for humans to understand and coordinate with. We show in chess and Go that regularizing search policies based on the KL divergence from an imitation-learned policy by applying Monte Carlo tree search produces policies that have higher human prediction accuracy and are stronger than the imitation policy. We then introduce a novel regret minimization algorithm that is regularized based on the KL divergence from an imitation-learned policy, and show that applying this algorithm to no-press Diplomacy yields a policy that maintains the same human prediction accuracy as imitation learning while being substantially stronger.

Prior AI successes in complex games have largely focused on settings with at most hundreds of actions at each decision point. In contrast, Diplomacy is a game with more than 10^20 possible actions per turn. Previous attempts to address games with large branching factors, such as Diplomacy, StarCraft, and Dota, used human data to bootstrap the policy or used handcrafted reward shaping. In this paper, we describe an algorithm for action exploration and equilibrium approximation in games with combinatorial action spaces. This algorithm simultaneously performs value iteration while learning a policy proposal network. A double oracle step is used to explore additional actions to add to the policy proposals. At each state, the target state value and policy for the model training are computed via an equilibrium search procedure. Using this algorithm, we train an agent, DORA, completely from scratch for a popular two-player variant of Diplomacy and show that it achieves superhuman performance. Additionally, we extend our methods to full-scale no-press Diplomacy and for the first time train an agent from scratch with no human data. We present evidence that this agent plays a strategy that is incompatible with human-data bootstrapped agents. This presents the first strong evidence of multiple equilibria in Diplomacy and suggests that self play alone may be insufficient for achieving superhuman performance in Diplomacy.

A common approach to solving physical-reasoning tasks is to train a value learner on example tasks. A limitation of such an approach is it requires learning about object dynamics solely from reward values assigned to the final state of a rollout of the environment. This study aims to address this limitation by augmenting the reward value with additional supervisory signals about object dynamics. Specifically,we define a distance measure between the trajectory of two target objects, and use this distance measure to characterize the similarity of two environment rollouts.We train the model to correctly rank rollouts according to this measure in addition to predicting the correct reward. Empirically, we find that this approach leads to substantial performance improvements on the PHYRE benchmark for physical reasoning: our approach obtains a new state-of-the-art on that benchmark.

Current large-scale auto-regressive language models (Radford et al., 2019; Liu et al., 2018; Graves, 2013) display impressive fluency and can generate convincing text. In this work we start by asking the question: Can the generations of these models be reliably distinguished from real text by statistical discriminators? We find experimentally that the answer is affirmative when we have access to the training data for the model, and guardedly affirmative even if we do not. This suggests that the auto-regressive models can be improved by incorporating the (globally normalized) discriminators into the generative process. We give a formalism for this using the Energy-Based Model framework, and show that it indeed improves the results of the generative models, measured both in terms of perplexity and in terms of human evaluation.

Prior AI breakthroughs in complex games have focused on either the purely adversarial or purely cooperative settings. In contrast, Diplomacy is a game of shifting alliances that involves both cooperation and competition. For this reason, Diplomacy has proven to be a formidable research challenge. In this paper we describe an agent for the no-press variant of Diplomacy that combines supervised learning on human data with one-step lookahead search via external regret minimization. External regret minimization techniques have been behind previous AI successes in adversarial games, most notably poker, but have not previously been shown to be successful in large-scale games involving cooperation. We show that our agent greatly exceeds the performance of past no-press Diplomacy bots, is unexploitable by expert humans, and achieves a rank of 23 out of 1,128 human players when playing anonymous games on a popular Diplomacy website. A primary goal for AI research is to develop agents that can act optimally in real-world multi-agent interactions (i.e., games). However, previous large-scale game AI results have focused on either purely competitive or purely cooperative settings. In contrast, real-world games, such as business negotiations, politics, and traffic navigation, involve a far more complex mixture of cooperation and competition. In such settings, the theoretical grounding for the techniques used in previous AI breakthroughs falls apart. In this paper we augment neural policies trained through imitation learning with regret minimization search techniques, and evaluate on the benchmark game of no-press Diplomacy.

The combination of deep reinforcement learning and search at both training and test time is a powerful paradigm that has led to a number of a successes in single-agent settings and perfect-information games, best exemplified by the success of AlphaZero. However, algorithms of this form have been unable to cope with imperfect-information games. This paper presents ReBeL, a general framework for self-play reinforcement learning and search for imperfect-information games. In the simpler setting of perfect-information games, ReBeL reduces to an algorithm similar to AlphaZero. Results show ReBeL leads to low exploitability in benchmark imperfect-information games and achieves superhuman performance in heads-up no-limit Texas hold'em poker, while using far less domain knowledge than any prior poker AI. We also prove that ReBeL converges to a Nash equilibrium in two-player zero-sum games in tabular settings.

Understanding and reasoning about physics is an important ability of intelligent agents. We develop the PHYRE benchmark for physical reasoning that contains a set of simple classical mechanics puzzles in a 2D physical environment. The benchmark is designed to encourage the development of learning algorithms that are sample-efficient and generalize well across puzzles. We test several modern learning algorithms on PHYRE and find that these algorithms fall short in solving the puzzles efficiently. We expect that PHYRE will encourage the development of novel sample-efficient agents that learn efficient but useful models of physics. For code and to play PHYRE for yourself, please visit https://player.phyre.ai.

Recent advances in generative modeling of text have demonstrated remarkable improvements in terms of fluency and coherency. In this work we investigate to which extent a machine can discriminate real from machine generated text. This is important in itself for automatic detection of computer generated stories, but can also serve as a tool for further improving text generation. We show that learning a dedicated scoring function to discriminate between real and fake text achieves higher precision than employing the likelihood of a generative model. The scoring functions generalize to other generators than those used for training as long as these generators have comparable model complexity and are trained on similar datasets.