Pierre-Yves Oudeyer

Ziang Xiao, Xingdi Yuan, Q. Vera Liao et al. · microsoft-research

Qualitative analysis of textual contents unpacks rich and valuable information by assigning labels to the data. However, this process is often labor-intensive, particularly when working with large datasets. While recent AI-based tools demonstrate utility, researchers may not have readily available AI resources and expertise, let alone be challenged by the limited generalizability of those task-specific models. In this study, we explored the use of large language models (LLMs) in supporting deductive coding, a major category of qualitative analysis where researchers use pre-determined codebooks to label the data into a fixed set of codes. Instead of training task-specific models, a pre-trained LLM could be used directly for various tasks without fine-tuning through prompt learning. Using a curiosity-driven questions coding task as a case study, we found, by combining GPT-3 with expert-drafted codebooks, our proposed approach achieved fair to substantial agreements with expert-coded results. We lay out challenges and opportunities in using LLMs to support qualitative coding and beyond.

Rania Abdelghani, Yen-Hsiang Wang, Xingdi Yuan et al. · microsoft-research

In order to train children's ability to ask curiosity-driven questions, previous research has explored designing specific exercises relying on providing semantic and linguistic cues to help formulate such questions. But despite showing pedagogical efficiency, this method is still limited as it relies on generating the said cues by hand, which can be a very costly process. In this context, we propose to leverage advances in the natural language processing field (NLP) and investigate the efficiency of using a large language model (LLM) for automating the production of the pedagogical content of a curious question-asking (QA) training. We study generating the said content using the "prompt-based" method that consists of explaining the task to the LLM in natural text. We evaluate the output using human experts annotations and comparisons with hand-generated content. Results suggested indeed the relevance and usefulness of this content. We also conduct a field study in primary school (75 children aged 9-10), where we evaluate children's QA performance when having this training. We compare 3 types of content : 1) hand-generated content that proposes "closed" cues leading to predefined questions; 2) GPT-3-generated content that proposes the same type of cues; 3) GPT-3-generated content that proposes "open" cues leading to several possible questions. We see a similar QA performance between the two "closed" trainings (showing the scalability of the approach using GPT-3), and a better one for participants with the "open" training. These results suggest the efficiency of using LLMs to support children in generating more curious questions, using a natural language prompting approach that affords usability by teachers and other users not specialists of AI techniques. Furthermore, results also show that open-ended content may be more suitable for training curious question-asking skills.

Xingdi Yuan, Tong Wang, Yen-Hsiang Wang et al. · microsoft-research

Large Language Models (LLMs) have in recent years demonstrated impressive prowess in natural language generation. A common practice to improve generation diversity is to sample multiple outputs from the model. However, there lacks a simple and robust way of selecting the best output from these stochastic samples. As a case study framed in the context of question generation, we propose two prompt-based approaches to selecting high-quality questions from a set of LLM-generated candidates. Our method works under the constraints of 1) a black-box (non-modifiable) question generation model and 2) lack of access to human-annotated references -- both of which are realistic limitations for real-world deployment of LLMs. With automatic as well as human evaluations, we empirically demonstrate that our approach can effectively select questions of higher qualities than greedy generation.

Iou-Jen Liu, Xingdi Yuan, Marc-Alexandre Côté et al. · microsoft-research

To solve difficult tasks, humans ask questions to acquire knowledge from external sources. In contrast, classical reinforcement learning agents lack such an ability and often resort to exploratory behavior. This is exacerbated as few present-day environments support querying for knowledge. In order to study how agents can be taught to query external knowledge via language, we first introduce two new environments: the grid-world-based Q-BabyAI and the text-based Q-TextWorld. In addition to physical interactions, an agent can query an external knowledge source specialized for these environments to gather information. Second, we propose the "Asking for Knowledge" (AFK) agent, which learns to generate language commands to query for meaningful knowledge that helps solve the tasks. AFK leverages a non-parametric memory, a pointer mechanism and an episodic exploration bonus to tackle (1) irrelevant information, (2) a large query language space, (3) delayed reward for making meaningful queries. Extensive experiments demonstrate that the AFK agent outperforms recent baselines on the challenging Q-BabyAI and Q-TextWorld environments.

Thomas Carta, Clément Romac, Thomas Wolf et al.

Recent works successfully leveraged Large Language Models' (LLM) abilities to capture abstract knowledge about world's physics to solve decision-making problems. Yet, the alignment between LLMs' knowledge and the environment can be wrong and limit functional competence due to lack of grounding. In this paper, we study an approach (named GLAM) to achieve this alignment through functional grounding: we consider an agent using an LLM as a policy that is progressively updated as the agent interacts with the environment, leveraging online Reinforcement Learning to improve its performance to solve goals. Using an interactive textual environment designed to study higher-level forms of functional grounding, and a set of spatial and navigation tasks, we study several scientific questions: 1) Can LLMs boost sample efficiency for online learning of various RL tasks? 2) How can it boost different forms of generalization? 3) What is the impact of online learning? We study these questions by functionally grounding several variants (size, architecture) of FLAN-T5.

Laetitia Teodorescu, Xingdi Yuan, Marc-Alexandre Côté et al. · microsoft-research

Building open-ended agents that can autonomously discover a diversity of behaviours is one of the long-standing goals of artificial intelligence. This challenge can be studied in the framework of autotelic RL agents, i.e. agents that learn by selecting and pursuing their own goals, self-organizing a learning curriculum. Recent work identified language as a key dimension of autotelic learning, in particular because it enables abstract goal sampling and guidance from social peers for hindsight relabelling. Within this perspective, we study the following open scientific questions: What is the impact of hindsight feedback from a social peer (e.g. selective vs. exhaustive)? How can the agent learn from very rare language goal examples in its experience replay? How can multiple forms of exploration be combined, and take advantage of easier goals as stepping stones to reach harder ones? To address these questions, we use ScienceWorld, a textual environment with rich abstract and combinatorial physics. We show the importance of selectivity from the social peer's feedback; that experience replay needs to over-sample examples of rare goals; and that following self-generated goal sequences where the agent's competence is intermediate leads to significant improvements in final performance.

Laetitia Teodorescu, Eric Yuan, Marc-Alexandre Côté et al. · microsoft-research

In this extended abstract we discuss the opportunities and challenges of studying intrinsically-motivated agents for exploration in textual environments. We argue that there is important synergy between text environments and autonomous agents. We identify key properties of text worlds that make them suitable for exploration by autonmous agents, namely, depth, breadth, progress niches and the ease of use of language goals; we identify drivers of exploration for such agents that are implementable in text worlds. We discuss the opportunities of using autonomous agents to make progress on text environment benchmarks. Finally we list some specific challenges that need to be overcome in this area.

Grgur Kovač, Masataka Sawayama, Rémy Portelas et al.

Large Language Models (LLMs) are often misleadingly recognized as having a personality or a set of values. We argue that an LLM can be seen as a superposition of perspectives with different values and personality traits. LLMs exhibit context-dependent values and personality traits that change based on the induced perspective (as opposed to humans, who tend to have more coherent values and personality traits across contexts). We introduce the concept of perspective controllability, which refers to a model's affordance to adopt various perspectives with differing values and personality traits. In our experiments, we use questionnaires from psychology (PVQ, VSM, IPIP) to study how exhibited values and personality traits change based on different perspectives. Through qualitative experiments, we show that LLMs express different values when those are (implicitly or explicitly) implied in the prompt, and that LLMs express different values even when those are not obviously implied (demonstrating their context-dependent nature). We then conduct quantitative experiments to study the controllability of different models (GPT-4, GPT-3.5, OpenAssistant, StableVicuna, StableLM), the effectiveness of various methods for inducing perspectives, and the smoothness of the models' drivability. We conclude by examining the broader implications of our work and outline a variety of associated scientific questions. The project website is available at https://sites.google.com/view/llm-superpositions .

Cédric Colas, Tristan Karch, Clément Moulin-Frier et al.

Building autonomous agents able to grow open-ended repertoires of skills across their lives is a fundamental goal of artificial intelligence (AI). A promising developmental approach recommends the design of intrinsically motivated agents that learn new skills by generating and pursuing their own goals - autotelic agents. But despite recent progress, existing algorithms still show serious limitations in terms of goal diversity, exploration, generalisation or skill composition. This perspective calls for the immersion of autotelic agents into rich socio-cultural worlds, an immensely important attribute of our environment that shapes human cognition but is mostly omitted in modern AI. Inspired by the seminal work of Vygotsky, we propose Vygotskian autotelic agents - agents able to internalise their interactions with others and turn them into cognitive tools. We focus on language and show how its structure and informational content may support the development of new cognitive functions in artificial agents as it does in humans. We justify the approach by uncovering several examples of new artificial cognitive functions emerging from interactions between language and embodiment in recent works at the intersection of deep reinforcement learning and natural language processing. Looking forward, we highlight future opportunities and challenges for Vygotskian Autotelic AI research, including the use of language models as cultural models supporting artificial cognitive development.

Thomas Carta, Pierre-Yves Oudeyer, Olivier Sigaud et al.

Reinforcement learning (RL) in long horizon and sparse reward tasks is notoriously difficult and requires a lot of training steps. A standard solution to speed up the process is to leverage additional reward signals, shaping it to better guide the learning process. In the context of language-conditioned RL, the abstraction and generalisation properties of the language input provide opportunities for more efficient ways of shaping the reward. In this paper, we leverage this idea and propose an automated reward shaping method where the agent extracts auxiliary objectives from the general language goal. These auxiliary objectives use a question generation (QG) and question answering (QA) system: they consist of questions leading the agent to try to reconstruct partial information about the global goal using its own trajectory. When it succeeds, it receives an intrinsic reward proportional to its confidence in its answer. This incentivizes the agent to generate trajectories which unambiguously explain various aspects of the general language goal. Our experimental study shows that this approach, which does not require engineer intervention to design the auxiliary objectives, improves sample efficiency by effectively directing exploration.

Grgur Kovač, Rémy Portelas, Peter Ford Dominey et al.

Developmental psychologists have long-established the importance of socio-cognitive abilities in human intelligence. These abilities enable us to enter, participate and benefit from human culture. AI research on social interactive agents mostly concerns the emergence of culture in a multi-agent setting (often without a strong grounding in developmental psychology). We argue that AI research should be informed by psychology and study socio-cognitive abilities enabling to enter a culture too. We discuss the theories of Michael Tomasello and Jerome Bruner to introduce some of their concepts to AI and outline key concepts and socio-cognitive abilities. We present The SocialAI school - a tool including a customizable parameterized uite of procedurally generated environments, which simplifies conducting experiments regarding those concepts. We show examples of such experiments with RL agents and Large Language Models. The main motivation of this work is to engage the AI community around the problem of social intelligence informed by developmental psychology, and to provide a tool to simplify first steps in this direction. Refer to the project website for code and additional information: https://sites.google.com/view/socialai-school.

Erwan Plantec, Gautier Hamon, Mayalen Etcheverry et al.

The design of complex self-organising systems producing life-like phenomena, such as the open-ended evolution of virtual creatures, is one of the main goals of artificial life. Lenia, a family of cellular automata (CA) generalizing Conway's Game of Life to continuous space, time and states, has attracted a lot of attention because of the wide diversity of self-organizing patterns it can generate. Among those, some spatially localized patterns (SLPs) resemble life-like artificial creatures and display complex behaviors. However, those creatures are found in only a small subspace of the Lenia parameter space and are not trivial to discover, necessitating advanced search algorithms. Furthermore, each of these creatures exist only in worlds governed by specific update rules and thus cannot interact in the same one. This paper proposes as mass-conservative extension of Lenia, called Flow Lenia, that solve both of these issues. We present experiments demonstrating its effectiveness in generating SLPs with complex behaviors and show that the update rule parameters can be optimized to generate SLPs showing behaviors of interest. Finally, we show that Flow Lenia enables the integration of the parameters of the CA update rules within the CA dynamics, making them dynamic and localized, allowing for multi-species simulations, with locally coherent update rules that define properties of the emerging creatures, and that can be mixed with neighbouring rules. We argue that this paves the way for the intrinsic evolution of self-organized artificial life forms within continuous CAs.

Eleni Nisioti, Mateo Mahaut, Pierre-Yves Oudeyer et al.

Human culture relies on innovation: our ability to continuously explore how existing elements can be combined to create new ones. Innovation is not solitary, it relies on collective search and accumulation. Reinforcement learning (RL) approaches commonly assume that fully-connected groups are best suited for innovation. However, human laboratory and field studies have shown that hierarchical innovation is more robustly achieved by dynamic social network structures. In dynamic settings, humans oscillate between innovating individually or in small clusters, and then sharing outcomes with others. To our knowledge, the role of social network structure on innovation has not been systematically studied in RL. Here, we use a multi-level problem setting (WordCraft), with three different innovation tasks to test the hypothesis that the social network structure affects the performance of distributed RL algorithms. We systematically design networks of DQNs sharing experiences from their replay buffers in varying structures (fully-connected, small world, dynamic, ring) and introduce a set of behavioral and mnemonic metrics that extend the classical reward-focused evaluation framework of RL. Comparing the level of innovation achieved by different social network structures across different tasks shows that, first, consistent with human findings, experience sharing within a dynamic structure achieves the highest level of innovation in tasks with a deceptive nature and large search spaces. Second, experience sharing is not as helpful when there is a single clear path to innovation. Third, the metrics we propose, can help understand the success of different social network structures on different tasks, with the diversity of experiences on an individual and group level lending crucial insights.

Jérémy Perez, Grgur Kovač, Corentin Léger et al.

As large language models (LLMs) start interacting with each other and generating an increasing amount of text online, it becomes crucial to better understand how information is transformed as it passes from one LLM to the next. While significant research has examined individual LLM behaviors, existing studies have largely overlooked the collective behaviors and information distortions arising from iterated LLM interactions. Small biases, negligible at the single output level, risk being amplified in iterated interactions, potentially leading the content to evolve towards attractor states. In a series of telephone game experiments, we apply a transmission chain design borrowed from the human cultural evolution literature: LLM agents iteratively receive, produce, and transmit texts from the previous to the next agent in the chain. By tracking the evolution of text toxicity, positivity, difficulty, and length across transmission chains, we uncover the existence of biases and attractors, and study their dependence on the initial text, the instructions, language model, and model size. For instance, we find that more open-ended instructions lead to stronger attraction effects compared to more constrained tasks. We also find that different text properties display different sensitivity to attraction effects, with toxicity leading to stronger attractors than length. These findings highlight the importance of accounting for multi-step transmission dynamics and represent a first step towards a more comprehensive understanding of LLM cultural dynamics.

Eleni Nisioti, Sebastian Risi, Ida Momennejad et al.

Human culture relies on collective innovation: our ability to continuously explore how existing elements in our environment can be combined to create new ones. Language is hypothesized to play a key role in human culture, driving individual cognitive capacities and shaping communication. Yet the majority of models of collective innovation assign no cognitive capacities or language abilities to agents. Here, we contribute a computational study of collective innovation where agents are Large Language Models (LLMs) that play Little Alchemy 2, a creative video game originally developed for humans that, as we argue, captures useful aspects of innovation landscapes not present in previous test-beds. We, first, study an LLM in isolation and discover that it exhibits both useful skills and crucial limitations. We, then, study groups of LLMs that share information related to their behaviour and focus on the effect of social connectivity on collective performance. In agreement with previous human and computational studies, we observe that groups with dynamic connectivity out-compete fully-connected groups. Our work reveals opportunities and challenges for future studies of collective innovation that are becoming increasingly relevant as Generative Artificial Intelligence algorithms and humans innovate alongside each other.

Julien Pourcel, Cédric Colas, Gaia Molinaro et al.

The ability to invent novel and interesting problems is a remarkable feature of human intelligence that drives innovation, art, and science. We propose a method that aims to automate this process by harnessing the power of state-of-the-art generative models to produce a diversity of challenging yet solvable problems, here in the context of Python programming puzzles. Inspired by the intrinsically motivated literature, Autotelic CodE Search (ACES) jointly optimizes for the diversity and difficulty of generated problems. We represent problems in a space of LLM-generated semantic descriptors describing the programming skills required to solve them (e.g. string manipulation, dynamic programming, etc.) and measure their difficulty empirically as a linearly decreasing function of the success rate of Llama-3-70B, a state-of-the-art LLM problem solver. ACES iteratively prompts a large language model to generate difficult problems achieving a diversity of target semantic descriptors (goal-directed exploration) using previously generated problems as in-context examples. ACES generates problems that are more diverse and more challenging than problems produced by baseline methods and three times more challenging than problems found in existing Python programming benchmarks on average across 11 state-of-the-art code LLMs.

Nicolas Yax, Pierre-Yves Oudeyer, Stefano Palminteri

In machine learning, contamination refers to situations where testing data leak into the training set. The issue is particularly relevant for the evaluation of the performance of Large Language Models (LLMs), which are generally trained on gargantuan, and generally opaque, corpora of text scraped from the world wide web. Developing tools to detect contamination is therefore crucial to be able to fairly and properly track the evolution of the performance of LLMs. To date, only a few recent studies have attempted to address the issue of quantifying and detecting contamination in short text sequences, such as those commonly found in benchmarks. However, these methods have limitations that can sometimes render them impractical. In the present paper, we introduce LogProber, a novel, efficient algorithm that we show to be able to detect contamination in a black box setting that tries to tackle some of these drawbacks by focusing on the familiarity with the question rather than the answer. Here, we explore the properties of the proposed method in comparison with concurrent approaches, identify its advantages and limitations, and illustrate how different forms of contamination can go undetected depending on the design of the detection algorithm.

Mayalen Etcheverry, Michael Levin, Clément Moulin-Frier et al.

Advances in bioengineering and biomedicine demand a deep understanding of the dynamic behavior of biological systems, ranging from protein pathways to complex cellular processes. Biological networks like gene regulatory networks and protein pathways are key drivers of embryogenesis and physiological processes. Comprehending their diverse behaviors is essential for tackling diseases, including cancer, as well as for engineering novel biological constructs. Despite the availability of extensive mathematical models represented in Systems Biology Markup Language (SBML), researchers face significant challenges in exploring the full spectrum of behaviors and optimizing interventions to efficiently shape those behaviors. Existing tools designed for simulation of biological network models are not tailored to facilitate interventions on network dynamics nor to facilitate automated discovery. Leveraging recent developments in machine learning (ML), this paper introduces SBMLtoODEjax, a lightweight library designed to seamlessly integrate SBML models with ML-supported pipelines, powered by JAX. SBMLtoODEjax facilitates the reuse and customization of SBML-based models, harnessing JAX's capabilities for efficient parallel simulations and optimization, with the aim to accelerate research in biological network analysis.

Tristan Karch, Yoann Lemesle, Romain Laroche et al.

In this paper, we investigate whether artificial agents can develop a shared language in an ecological setting where communication relies on a sensory-motor channel. To this end, we introduce the Graphical Referential Game (GREG) where a speaker must produce a graphical utterance to name a visual referent object while a listener has to select the corresponding object among distractor referents, given the delivered message. The utterances are drawing images produced using dynamical motor primitives combined with a sketching library. To tackle GREG we present CURVES: a multimodal contrastive deep learning mechanism that represents the energy (alignment) between named referents and utterances generated through gradient ascent on the learned energy landscape. We demonstrate that CURVES not only succeeds at solving the GREG but also enables agents to self-organize a language that generalizes to feature compositions never seen during training. In addition to evaluating the communication performance of our approach, we also explore the structure of the emerging language. Specifically, we show that the resulting language forms a coherent lexicon shared between agents and that basic compositional rules on the graphical productions could not explain the compositional generalization.

Olivier Sigaud, Gianluca Baldassarre, Cedric Colas et al.

A lot of recent machine learning research papers have ``open-ended learning'' in their title. But very few of them attempt to define what they mean when using the term. Even worse, when looking more closely there seems to be no consensus on what distinguishes open-ended learning from related concepts such as continual learning, lifelong learning or autotelic learning. In this paper, we contribute to fixing this situation. After illustrating the genealogy of the concept and more recent perspectives about what it truly means, we outline that open-ended learning is generally conceived as a composite notion encompassing a set of diverse properties. In contrast with previous approaches, we propose to isolate a key elementary property of open-ended processes, which is to produce elements from time to time (e.g., observations, options, reward functions, and goals), over an infinite horizon, that are considered novel from an observer's perspective. From there, we build the notion of open-ended learning problems and focus in particular on the subset of open-ended goal-conditioned reinforcement learning problems in which agents can learn a growing repertoire of goal-driven skills. Finally, we highlight the work that remains to be performed to fill the gap between our elementary definition and the more involved notions of open-ended learning that developmental AI researchers may have in mind.

Jérémy Perez, Corentin Léger, Marcela Ovando-Tellez et al.

Research in cultural evolution aims at providing causal explanations for the change of culture over time. Over the past decades, this field has generated an important body of knowledge, using experimental, historical, and computational methods. While computational models have been very successful at generating testable hypotheses about the effects of several factors, such as population structure or transmission biases, some phenomena have so far been more complex to capture using agent-based and formal models. This is in particular the case for the effect of the transformations of social information induced by evolved cognitive mechanisms. We here propose that leveraging the capacity of Large Language Models (LLMs) to mimic human behavior may be fruitful to address this gap. On top of being an useful approximation of human cultural dynamics, multi-agents models featuring generative agents are also important to study for their own sake. Indeed, as artificial agents are bound to participate more and more to the evolution of culture, it is crucial to better understand the dynamics of machine-generated cultural evolution. We here present a framework for simulating cultural evolution in populations of LLMs, allowing the manipulation of variables known to be important in cultural evolution, such as network structure, personality, and the way social information is aggregated and transformed. The software we developed for conducting these simulations is open-source and features an intuitive user-interface, which we hope will help to build bridges between the fields of cultural evolution and generative artificial intelligence.

Julien Pourcel, Cédric Colas, Pierre-Yves Oudeyer

Many program synthesis tasks prove too challenging for even state-of-the-art language models to solve in single attempts. Search-based evolutionary methods offer a promising alternative by exploring solution spaces iteratively, but their effectiveness remain limited by the fixed capabilities of the underlying generative model. We propose SOAR, a method that learns program synthesis by integrating language models into a self-improving evolutionary loop. SOAR alternates between (1) an evolutionary search that uses an LLM to sample and refine candidate solutions, and (2) a hindsight learning phase that converts search attempts into valid problem-solution pairs used to fine-tune the LLM's sampling and refinement capabilities\, -- \,enabling increasingly effective search in subsequent iterations. On the challenging ARC-AGI benchmark, SOAR achieves significant performance gains across model scales and iterations, leveraging positive transfer between the sampling and refinement finetuning tasks. These improvements carry over to test-time adaptation, enabling SOAR to solve 52\% of the public test set. Our code is open-sourced at: https://github.com/flowersteam/SOAR

Nicolas Yax, Pierre-Yves Oudeyer, Stefano Palminteri

This paper introduces PhyloLM, a method adapting phylogenetic algorithms to Large Language Models (LLMs) to explore whether and how they relate to each other and to predict their performance characteristics. Our method calculates a phylogenetic distance metric based on the similarity of LLMs' output. The resulting metric is then used to construct dendrograms, which satisfactorily capture known relationships across a set of 111 open-source and 45 closed models. Furthermore, our phylogenetic distance predicts performance in standard benchmarks, thus demonstrating its functional validity and paving the way for a time and cost-effective estimation of LLM capabilities. To sum up, by translating population genetic concepts to machine learning, we propose and validate a tool to evaluate LLM development, relationships and capabilities, even in the absence of transparent training information.

Tristan Karch, Laetitia Teodorescu, Katja Hofmann et al.

Language is an interface to the outside world. In order for embodied agents to use it, language must be grounded in other, sensorimotor modalities. While there is an extended literature studying how machines can learn grounded language, the topic of how to learn spatio-temporal linguistic concepts is still largely uncharted. To make progress in this direction, we here introduce a novel spatio-temporal language grounding task where the goal is to learn the meaning of spatio-temporal descriptions of behavioral traces of an embodied agent. This is achieved by training a truth function that predicts if a description matches a given history of observations. The descriptions involve time-extended predicates in past and present tense as well as spatio-temporal references to objects in the scene. To study the role of architectural biases in this task, we train several models including multimodal Transformer architectures; the latter implement different attention computations between words and objects across space and time. We test models on two classes of generalization: 1) generalization to randomly held-out sentences; 2) generalization to grammar primitives. We observe that maintaining object identity in the attention computation of our Transformers is instrumental to achieving good performance on generalization overall, and that summarizing object traces in a single token has little influence on performance. We then discuss how this opens new perspectives for language-guided autonomous embodied agents. We also release our code under open-source license as well as pretrained models and datasets to encourage the wider community to build upon and extend our work in the future.

Clément Romac, Rémy Portelas, Katja Hofmann et al.

Training autonomous agents able to generalize to multiple tasks is a key target of Deep Reinforcement Learning (DRL) research. In parallel to improving DRL algorithms themselves, Automatic Curriculum Learning (ACL) study how teacher algorithms can train DRL agents more efficiently by adapting task selection to their evolving abilities. While multiple standard benchmarks exist to compare DRL agents, there is currently no such thing for ACL algorithms. Thus, comparing existing approaches is difficult, as too many experimental parameters differ from paper to paper. In this work, we identify several key challenges faced by ACL algorithms. Based on these, we present TeachMyAgent (TA), a benchmark of current ACL algorithms leveraging procedural task generation. It includes 1) challenge-specific unit-tests using variants of a procedural Box2D bipedal walker environment, and 2) a new procedural Parkour environment combining most ACL challenges, making it ideal for global performance assessment. We then use TeachMyAgent to conduct a comparative study of representative existing approaches, showcasing the competitiveness of some ACL algorithms that do not use expert knowledge. We also show that the Parkour environment remains an open problem. We open-source our environments, all studied ACL algorithms (collected from open-source code or re-implemented), and DRL students in a Python package available at https://github.com/flowersteam/TeachMyAgent.

Grgur Kovač, Adrien Laversanne-Finot, Pierre-Yves Oudeyer

Designing agents, capable of learning autonomously a wide range of skills is critical in order to increase the scope of reinforcement learning. It will both increase the diversity of learned skills and reduce the burden of manually designing reward functions for each skill. Self-supervised agents, setting their own goals, and trying to maximize the diversity of those goals have shown great promise towards this end. However, a currently known limitation of agents trying to maximize the diversity of sampled goals is that they tend to get attracted to noise or more generally to parts of the environments that cannot be controlled (distractors). When agents have access to predefined goal features or expert knowledge, absolute Learning Progress (ALP) provides a way to distinguish between regions that can be controlled and those that cannot. However, those methods often fall short when the agents are only provided with raw sensory inputs such as images. In this work we extend those concepts to unsupervised image-based goal exploration. We propose a framework that allows agents to autonomously identify and ignore noisy distracting regions while searching for novelty in the learnable regions to both improve overall performance and avoid catastrophic forgetting. Our framework can be combined with any state-of-the-art novelty seeking goal exploration approaches. We construct a rich 3D image based environment with distractors. Experiments on this environment show that agents using our framework successfully identify interesting regions of the environment, resulting in drastically improved performances. The source code is available at https://sites.google.com/view/grimgep.

Rémy Portelas, Cédric Colas, Katja Hofmann et al.

We consider the problem of how a teacher algorithm can enable an unknown Deep Reinforcement Learning (DRL) student to become good at a skill over a wide range of diverse environments. To do so, we study how a teacher algorithm can learn to generate a learning curriculum, whereby it sequentially samples parameters controlling a stochastic procedural generation of environments. Because it does not initially know the capacities of its student, a key challenge for the teacher is to discover which environments are easy, difficult or unlearnable, and in what order to propose them to maximize the efficiency of learning over the learnable ones. To achieve this, this problem is transformed into a surrogate continuous bandit problem where the teacher samples environments in order to maximize absolute learning progress of its student. We present a new algorithm modeling absolute learning progress with Gaussian mixture models (ALP-GMM). We also adapt existing algorithms and provide a complete study in the context of DRL. Using parameterized variants of the BipedalWalker environment, we study their efficiency to personalize a learning curriculum for different learners (embodiments), their robustness to the ratio of learnable/unlearnable environments, and their scalability to non-linear and high-dimensional parameter spaces. Videos and code are available at https://github.com/flowersteam/teachDeepRL.

Cédric Colas, Olivier Sigaud, Pierre-Yves Oudeyer

Consistently checking the statistical significance of experimental results is one of the mandatory methodological steps to address the so-called "reproducibility crisis" in deep reinforcement learning. In this tutorial paper, we explain how the number of random seeds relates to the probabilities of statistical errors. For both the t-test and the bootstrap confidence interval test, we recall theoretical guidelines to determine the number of random seeds one should use to provide a statistically significant comparison of the performance of two algorithms. Finally, we discuss the influence of deviations from the assumptions usually made by statistical tests. We show that they can lead to inaccurate evaluations of statistical errors and provide guidelines to counter these negative effects. We make our code available to perform the tests.

Cédric Colas, Olivier Sigaud, Pierre-Yves Oudeyer

In continuous action domains, standard deep reinforcement learning algorithms like DDPG suffer from inefficient exploration when facing sparse or deceptive reward problems. Conversely, evolutionary and developmental methods focusing on exploration like Novelty Search, Quality-Diversity or Goal Exploration Processes explore more robustly but are less efficient at fine-tuning policies using gradient descent. In this paper, we present the GEP-PG approach, taking the best of both worlds by sequentially combining a Goal Exploration Process and two variants of DDPG. We study the learning performance of these components and their combination on a low dimensional deceptive reward problem and on the larger Half-Cheetah benchmark. We show that DDPG fails on the former and that GEP-PG improves over the best DDPG variant in both environments. Supplementary videos and discussion can be found at http://frama.link/gep_pg, the code at http://github.com/flowersteam/geppg.

Chloé Desvaux, Rania Abdelghani, Pierre-Yves Oudeyer et al.

This chapter examines the relationship between curiosity and metacognition as critical drivers of autonomous and self-regulated learning. We synthesize recent research to propose a unified framework integrating behavioral, computational, and psychoeducational dimensions, arguing that curiosity, i.e. the intrinsic drive to acquire new knowledge, relies fundamentally on metacognitive monitoring and control. From an educational perspective, we evaluate interventions designed to enhance curiosity in classroom settings. While promising, our review indicates that these interventions yield mixed results, often proving differentially effective for struggling learners, thereby underscoring the necessity for approaches tailored to individual profiles. Finally, we address the paradigm shift introduced by Generative AI. While Large Language Models (LLMs) offer unprecedented scalability for personalized inquiry, we argue that their default interaction modes pose significant risks to the dynamics of curiosity-driven learning. To mitigate these challenges, we review strategies to transform AI from a potential cognitive shortcut into a powerful partner for sustained epistemic development.

Grgur Kovač, Rémy Portelas, Masataka Sawayama et al.

The standard way to study Large Language Models (LLMs) with benchmarks or psychology questionnaires is to provide many different queries from similar minimal contexts (e.g. multiple choice questions). However, due to LLMs' highly context-dependent nature, conclusions from such minimal-context evaluations may be little informative about the model's behavior in deployment (where it will be exposed to many new contexts). We argue that context-dependence (specifically, value stability) should be studied as a specific property of LLMs and used as another dimension of LLM comparison (alongside others such as cognitive abilities, knowledge, or model size). We present a case-study on the stability of value expression over different contexts (simulated conversations on different topics) as measured using a standard psychology questionnaire (PVQ) and on behavioral downstream tasks. Reusing methods from psychology, we study Rank-order stability on the population (interpersonal) level, and Ipsative stability on the individual (intrapersonal) level. We consider two settings (with and without instructing LLMs to simulate particular personas), two simulated populations, and three downstream tasks. We observe consistent trends in the stability of models and model families - Mixtral, Mistral, GPT-3.5 and Qwen families are more stable than LLaMa-2 and Phi. The consistency of these trends implies that some models exhibit higher value stability than others, and that stability can be estimated with the set of introduced methodological tools. When instructed to simulate particular personas, LLMs exhibit low Rank-order stability, which further diminishes with conversation length. This highlights the need for future research on LLMs that coherently simulate different personas. This paper provides a foundational step in that direction, and, to our knowledge, it is the first study of value stability in LLMs.

Gautier Hamon, Mayalen Etcheverry, Bert Wang-Chak Chan et al.

The research field of Artificial Life studies how life-like phenomena such as autopoiesis, agency, or self-regulation can self-organize in computer simulations. In cellular automata (CA), a key open-question has been whether it it is possible to find environment rules that self-organize robust "individuals" from an initial state with no prior existence of things like "bodies", "brain", "perception" or "action". In this paper, we leverage recent advances in machine learning, combining algorithms for diversity search, curriculum learning and gradient descent, to automate the search of such "individuals", i.e. localized structures that move around with the ability to react in a coherent manner to external obstacles and maintain their integrity, hence primitive forms of sensorimotor agency. We show that this approach enables to find systematically environmental conditions in CA leading to self-organization of such basic forms of agency. Through multiple experiments, we show that the discovered agents have surprisingly robust capabilities to move, maintain their body integrity and navigate among various obstacles. They also show strong generalization abilities, with robustness to changes of scale, random updates or perturbations from the environment not seen during training. We discuss how this approach opens new perspectives in AI and synthetic bioengineering.

Junior Cedric Tonga, Benjamin Clement, Pierre-Yves Oudeyer

The automatic generation of hints by Large Language Models (LLMs) within Intelligent Tutoring Systems (ITSs) has shown potential to enhance student learning. However, generating pedagogically sound hints that address student misconceptions and adhere to specific educational objectives remains challenging. This work explores using LLMs (GPT-4o and Llama-3-8B-instruct) as teachers to generate effective hints for students simulated through LLMs (GPT-3.5-turbo, Llama-3-8B-Instruct, or Mistral-7B-instruct-v0.3) tackling math exercises designed for human high-school students, and designed using cognitive science principles. We present here the study of several dimensions: 1) identifying error patterns made by simulated students on secondary-level math exercises; 2) developing various prompts for GPT-4o as a teacher and evaluating their effectiveness in generating hints that enable simulated students to self-correct; and 3) testing the best-performing prompts, based on their ability to produce relevant hints and facilitate error correction, with Llama-3-8B-Instruct as the teacher, allowing for a performance comparison with GPT-4o. The results show that model errors increase with higher temperature settings. Notably, when hints are generated by GPT-4o, the most effective prompts include prompts tailored to specific errors as well as prompts providing general hints based on common mathematical errors. Interestingly, Llama-3-8B-Instruct as a teacher showed better overall performance than GPT-4o. Also the problem-solving and response revision capabilities of the LLMs as students, particularly GPT-3.5-turbo, improved significantly after receiving hints, especially at lower temperature settings. However, models like Mistral-7B-Instruct demonstrated a decline in performance as the temperature increased.

Loris Gaven, Clement Romac, Thomas Carta et al.

The past years have seen Large Language Models (LLMs) strive not only as generative models but also as agents solving textual sequential decision-making tasks. When facing complex environments where their zero-shot abilities are insufficient, recent work showed online Reinforcement Learning (RL) could be used for the LLM agent to discover and learn efficient strategies interactively. However, most prior work sticks to on-policy algorithms, which greatly reduces the scope of methods such agents could use for both exploration and exploitation, such as experience replay and hindsight relabeling. Yet, such methods may be key for LLM learning agents, and in particular when designing autonomous intrinsically motivated agents sampling and pursuing their own goals (i.e. autotelic agents). This paper presents and studies an adaptation of Soft Actor-Critic and hindsight relabeling to LLM agents. Our method not only paves the path towards autotelic LLM agents that learn online but can also outperform on-policy methods in more classic multi-goal RL environments.

Loris Gaven, Thomas Carta, Clément Romac et al.

Open-ended learning agents must efficiently prioritize goals in vast possibility spaces, focusing on those that maximize learning progress (LP). When such autotelic exploration is achieved by LLM agents trained with online RL in high-dimensional and evolving goal spaces, a key challenge for LP prediction is modeling one's own competence, a form of metacognitive monitoring. Traditional approaches either require extensive sampling or rely on brittle expert-defined goal groupings. We introduce MAGELLAN, a metacognitive framework that lets LLM agents learn to predict their competence and LP online. By capturing semantic relationships between goals, MAGELLAN enables sample-efficient LP estimation and dynamic adaptation to evolving goal spaces through generalization. In an interactive learning environment, we show that MAGELLAN improves LP prediction efficiency and goal prioritization, being the only method allowing the agent to fully master a large and evolving goal space. These results demonstrate how augmenting LLM agents with a metacognitive ability for LP predictions can effectively scale curriculum learning to open-ended goal spaces.

Mohamed Salim Aissi, Clement Romac, Thomas Carta et al.

Reinforcement learning (RL) is a promising approach for aligning large language models (LLMs) knowledge with sequential decision-making tasks. However, few studies have thoroughly investigated the impact on LLM agents capabilities of fine-tuning them with RL in a specific environment. In this paper, we propose a novel framework to analyze the sensitivity of LLMs to prompt formulations following RL training in a textual environment. Our findings reveal that the performance of LLMs degrades when faced with prompt formulations different from those used during the RL training phase. Besides, we analyze the source of this sensitivity by examining the model's internal representations and salient tokens. Finally, we propose to use a contrastive loss to mitigate this sensitivity and improve the robustness and generalization capabilities of LLMs.

Sina Khajehabdollahi, Gautier Hamon, Marko Cvjetko et al.

Discovering diverse visual patterns in continuous cellular automata (CA) is challenging due to the vastness and redundancy of high-dimensional behavioral spaces. Traditional exploration methods like Novelty Search (NS) expand locally by mutating known novel solutions but often plateau when local novelty is exhausted, failing to reach distant, unexplored regions. We introduce Expedition and Expansion (E&E), a hybrid strategy where exploration alternates between local novelty-driven expansions and goal-directed expeditions. During expeditions, E&E leverages a Vision-Language Model (VLM) to generate linguistic goals--descriptions of interesting but hypothetical patterns that drive exploration toward uncharted regions. By operating in semantic spaces that align with human perception, E&E both evaluates novelty and generates goals in conceptually meaningful ways, enhancing the interpretability and relevance of discovered behaviors. Tested on Flow Lenia, a continuous CA known for its rich, emergent behaviors, E&E consistently uncovers more diverse solutions than existing exploration methods. A genealogical analysis further reveals that solutions originating from expeditions disproportionately influence long-term exploration, unlocking new behavioral niches that serve as stepping stones for subsequent search. These findings highlight E&E's capacity to break through local novelty boundaries and explore behavioral landscapes in human-aligned, interpretable ways, offering a promising template for open-ended exploration in artificial life and beyond.

Erwan Plantec, Gautier Hamon, Mayalen Etcheverry et al.

Central to the artificial life endeavour is the creation of artificial systems spontaneously generating properties found in the living world such as autopoiesis, self-replication, evolution and open-endedness. While numerous models and paradigms have been proposed, cellular automata (CA) have taken a very important place in the field notably as they enable the study of phenomenons like self-reproduction and autopoiesis. Continuous CA like Lenia have been showed to produce life-like patterns reminiscent, on an aesthetic and ontological point of view, of biological organisms we call creatures. We propose in this paper Flow-Lenia, a mass conservative extension of Lenia. We present experiments demonstrating its effectiveness in generating spatially-localized patters (SLPs) with complex behaviors and show that the update rule parameters can be optimized to generate complex creatures showing behaviors of interest. Furthermore, we show that Flow-Lenia allows us to embed the parameters of the model, defining the properties of the emerging patterns, within its own dynamics thus allowing for multispecies simulations. By using the evolutionary activity framework as well as other metrics, we shed light on the emergent evolutionary dynamics taking place in this system.

Thomas Michel, Marko Cvjetko, Gautier Hamon et al.

We present a method for the automated discovery of system-level dynamics in Flow-Lenia--a continuous cellular automaton (CA) with mass conservation and parameter localization-using a curiosity--driven AI scientist. This method aims to uncover processes leading to self-organization of evolutionary and ecosystemic dynamics in CAs. We build on previous work which uses diversity search algorithms in Lenia to find self-organized individual patterns, and extend it to large environments that support distinct interacting patterns. We adapt Intrinsically Motivated Goal Exploration Processes (IMGEPs) to drive exploration of diverse Flow-Lenia environments using simulation-wide metrics, such as evolutionary activity, compression-based complexity, and multi-scale entropy. We test our method in two experiments, showcasing its ability to illuminate significantly more diverse dynamics compared to random search. We show qualitative results illustrating how ecosystemic simulations enable self-organization of complex collective behaviors not captured by previous individual pattern search and analysis. We complement automated discovery with an interactive exploration tool, creating an effective human-AI collaborative workflow for scientific investigation. Though demonstrated specifically with Flow-Lenia, this methodology provides a framework potentially applicable to other parameterizable complex systems where understanding emergent collective properties is of interest.

Grgur Kovač, Jérémy Perez, Rémy Portelas et al.

Large language models (LLMs) are increasingly used in the creation of online content, creating feedback loops as subsequent generations of models will be trained on this synthetic data. Such loops were shown to lead to distribution shifts - models misrepresenting the true underlying distributions of human data (also called model collapse). However, how human data properties affect such shifts remains poorly understood. In this paper, we provide the first empirical examination of the effect of such properties on the outcome of recursive training. We first confirm that using different human datasets leads to distribution shifts of different magnitudes. Through exhaustive manipulation of dataset properties combined with regression analyses, we then identify a set of properties predicting distribution shift magnitudes. Lexical diversity is found to amplify these shifts, while semantic diversity and data quality mitigate them. Furthermore, we find that these influences are highly modular: data scrapped from a given internet domain has little influence on the content generated for another domain. Finally, experiments on political bias reveal that human data properties affect whether the initial bias will be amplified or reduced. Overall, our results portray a novel view, where different parts of internet may undergo different types of distribution shift.

Thomas Carta, Clément Romac, Loris Gaven et al.

Open-ended AI agents need to be able to learn efficiently goals of increasing complexity, abstraction and heterogeneity over their lifetime. Beyond sampling efficiently their own goals, autotelic agents specifically need to be able to keep the growing complexity of goals under control, limiting the associated growth in sample and computational complexity. To adress this challenge, recent approaches have leveraged hierarchical reinforcement learning (HRL) and language, capitalizing on its compositional and combinatorial generalization capabilities to acquire temporally extended reusable behaviours. Existing approaches use expert defined spaces of subgoals over which they instantiate a hierarchy, and often assume pre-trained associated low-level policies. Such designs are inadequate in open-ended scenarios, where goal spaces naturally diversify across a broad spectrum of difficulties. We introduce HERAKLES, a framework that enables a two-level hierarchical autotelic agent to continuously compile mastered goals into the low-level policy, executed by a small, fast neural network, dynamically expanding the set of subgoals available to the high-level policy. We train a Large Language Model (LLM) to serve as the high-level controller, exploiting its strengths in goal decomposition and generalization to operate effectively over this evolving subgoal space. We evaluate HERAKLES in the open-ended Crafter environment and show that it scales effectively with goal complexity, improves sample efficiency through skill compilation, and enables the agent to adapt robustly to novel challenges over time.

Benjamin Clément, Hélène Sauzéon, Didier Roy et al.

Large class sizes challenge personalized learning in schools, prompting the use of educational technologies such as intelligent tutoring systems. To address this, we present an AI-driven personalization system, called ZPDES, based on the Learning Progress Hypothesis - modeling curiosity-driven learning - and multi-armed bandit techniques. It sequences exercises that maximize learning progress for each student. While previous studies demonstrated its efficacy in enhancing learning compared to hand-made curricula, its impact on student motivation remained unexplored. Furthermore, ZPDES previously lacked features allowing student choice, a limitation in agency that conflicts with its foundation on models of curiosity-driven learning. This study investigates how integrating choice, as a gamification element unrelated to exercise difficulty, affects both learning outcomes and motivation. We conducted an extensive field study (265 7-8 years old children, RCT design), comparing ZPDES with and without choice against a hand-designed curriculum. Results show that ZPDES improves both learning performance and the learning experience. Moreover adding choice to ZPDES enhances intrinsic motivation and further strengthens its learning benefits. In contrast, incorporating choice into a fixed, linear curriculum negatively impacts learning outcomes. These findings highlight that the intrinsic motivation elicited by choice (gamification) is beneficial only when paired with an adaptive personalized learning system. This insight is critical as gamified features become increasingly prevalent in educational technologies.

Cédric Colas, Laetitia Teodorescu, Pierre-Yves Oudeyer et al.

Humans learn to master open-ended repertoires of skills by imagining and practicing their own goals. This autotelic learning process, literally the pursuit of self-generated (auto) goals (telos), becomes more and more open-ended as the goals become more diverse, abstract and creative. The resulting exploration of the space of possible skills is supported by an inter-individual exploration: goal representations are culturally evolved and transmitted across individuals, in particular using language. Current artificial agents mostly rely on predefined goal representations corresponding to goal spaces that are either bounded (e.g. list of instructions), or unbounded (e.g. the space of possible visual inputs) but are rarely endowed with the ability to reshape their goal representations, to form new abstractions or to imagine creative goals. In this paper, we introduce a language model augmented autotelic agent (LMA3) that leverages a pretrained language model (LM) to support the representation, generation and learning of diverse, abstract, human-relevant goals. The LM is used as an imperfect model of human cultural transmission; an attempt to capture aspects of humans' common-sense, intuitive physics and overall interests. Specifically, it supports three key components of the autotelic architecture: 1)~a relabeler that describes the goals achieved in the agent's trajectories, 2)~a goal generator that suggests new high-level goals along with their decomposition into subgoals the agent already masters, and 3)~reward functions for each of these goals. Without relying on any hand-coded goal representations, reward functions or curriculum, we show that LMA3 agents learn to master a large diversity of skills in a task-agnostic text-based environment.

Yoann Lemesle, Masataka Sawayama, Guillermo Valle-Perez et al.

Humans show language-biased image recognition for a word-embedded image, known as picture-word interference. Such interference depends on hierarchical semantic categories and reflects that human language processing highly interacts with visual processing. Similar to humans, recent artificial models jointly trained on texts and images, e.g., OpenAI CLIP, show language-biased image classification. Exploring whether the bias leads to interference similar to those observed in humans can contribute to understanding how much the model acquires hierarchical semantic representations from joint learning of language and vision. The present study introduces methodological tools from the cognitive science literature to assess the biases of artificial models. Specifically, we introduce a benchmark task to test whether words superimposed on images can distort the image classification across different category levels and, if it can, whether the perturbation is due to the shared semantic representation between language and vision. Our dataset is a set of word-embedded images and consists of a mixture of natural image datasets and hierarchical word labels with superordinate/basic category levels. Using this benchmark test, we evaluate the CLIP model. We show that presenting words distorts the image classification by the model across different category levels, but the effect does not depend on the semantic relationship between images and embedded words. This suggests that the semantic word representation in the CLIP visual processing is not shared with the image representation, although the word representation strongly dominates for word-embedded images.

Paul Barde, Tristan Karch, Derek Nowrouzezahrai et al.

We are interested in interactive agents that learn to coordinate, namely, a $builder$ -- which performs actions but ignores the goal of the task, i.e. has no access to rewards -- and an $architect$ which guides the builder towards the goal of the task. We define and explore a formal setting where artificial agents are equipped with mechanisms that allow them to simultaneously learn a task while at the same time evolving a shared communication protocol. Ideally, such learning should only rely on high-level communication priors and be able to handle a large variety of tasks and meanings while deriving communication protocols that can be reused across tasks. We present the Architect-Builder Problem (ABP): an asymmetrical setting in which an architect must learn to guide a builder towards constructing a specific structure. The architect knows the target structure but cannot act in the environment and can only send arbitrary messages to the builder. The builder on the other hand can act in the environment, but receives no rewards nor has any knowledge about the task, and must learn to solve it relying only on the messages sent by the architect. Crucially, the meaning of messages is initially not defined nor shared between the agents but must be negotiated throughout learning. Under these constraints, we propose Architect-Builder Iterated Guiding (ABIG), a solution to ABP where the architect leverages a learned model of the builder to guide it while the builder uses self-imitation learning to reinforce its guided behavior. We analyze the key learning mechanisms of ABIG and test it in 2D tasks involving grasping cubes, placing them at a given location, or building various shapes. ABIG results in a low-level, high-frequency, guiding communication protocol that not only enables an architect-builder pair to solve the task at hand, but that can also generalize to unseen tasks.

Jessy Ceha, Edith Law, Dana Kulić et al.

With advances in artificial intelligence, research is increasingly exploring the potential functions that social robots can play in education. As teachers are a critical stakeholder in the use and application of educational technologies, we conducted a study to understand teachers' perspectives on how a social robot could support a variety of learning activities in the classroom. Through interviews, robot puppeteering, and group brainstorming sessions with five elementary and middle school teachers from a local school in Canada, we take a socio-technical perspective to conceptualize possible robot functions and behaviours, and the effects they may have on the current way learning activities are designed, planned, and executed. Overall, the teachers responded positively to the idea of introducing a social robot as a technological tool for learning activities, envisioning differences in usage for teacher-robot and student-robot interactions. Further, Engeström's Activity System Model -- a framework for analyzing human needs, tasks, and outcomes -- illustrated a number of tensions associated with learning activities in the classroom. We discuss the fine-grained robot functions and behaviours conceived by teachers, and how they address the current tensions -- providing suggestions for improving the design of social robots for learning activities.

Grgur Kovač, Rémy Portelas, Katja Hofmann et al.

Building embodied autonomous agents capable of participating in social interactions with humans is one of the main challenges in AI. Within the Deep Reinforcement Learning (DRL) field, this objective motivated multiple works on embodied language use. However, current approaches focus on language as a communication tool in very simplified and non-diverse social situations: the "naturalness" of language is reduced to the concept of high vocabulary size and variability. In this paper, we argue that aiming towards human-level AI requires a broader set of key social skills: 1) language use in complex and variable social contexts; 2) beyond language, complex embodied communication in multimodal settings within constantly evolving social worlds. We explain how concepts from cognitive sciences could help AI to draw a roadmap towards human-like intelligence, with a focus on its social dimensions. As a first step, we propose to expand current research to a broader set of core social skills. To do this, we present SocialAI, a benchmark to assess the acquisition of social skills of DRL agents using multiple grid-world environments featuring other (scripted) social agents. We then study the limits of a recent SOTA DRL approach when tested on SocialAI and discuss important next steps towards proficient social agents. Videos and code are available at https://sites.google.com/view/socialai.

Maxime Gasse, Damien Grasset, Guillaume Gaudron et al.

Learning efficiently a causal model of the environment is a key challenge of model-based RL agents operating in POMDPs. We consider here a scenario where the learning agent has the ability to collect online experiences through direct interactions with the environment (interventional data), but has also access to a large collection of offline experiences, obtained by observing another agent interacting with the environment (observational data). A key ingredient, that makes this situation non-trivial, is that we allow the observed agent to interact with the environment based on hidden information, which is not observed by the learning agent. We then ask the following questions: can the online and offline experiences be safely combined for learning a causal model ? And can we expect the offline experiences to improve the agent's performances ? To answer these questions, we import ideas from the well-established causal framework of do-calculus, and we express model-based reinforcement learning as a causal inference problem. Then, we propose a general yet simple methodology for leveraging offline data during learning. In a nutshell, the method relies on learning a latent-based causal transition model that explains both the interventional and observational regimes, and then using the recovered latent variable to infer the standard POMDP transition model via deconfounding. We prove our method is correct and efficient in the sense that it attains better generalization guarantees due to the offline data (in the asymptotic case), and we illustrate its effectiveness empirically on synthetic toy problems. Our contribution aims at bridging the gap between the fields of reinforcement learning and causality.

Guillermo Valle-Pérez, Gustav Eje Henter, Jonas Beskow et al.

Dance requires skillful composition of complex movements that follow rhythmic, tonal and timbral features of music. Formally, generating dance conditioned on a piece of music can be expressed as a problem of modelling a high-dimensional continuous motion signal, conditioned on an audio signal. In this work we make two contributions to tackle this problem. First, we present a novel probabilistic autoregressive architecture that models the distribution over future poses with a normalizing flow conditioned on previous poses as well as music context, using a multimodal transformer encoder. Second, we introduce the currently largest 3D dance-motion dataset, obtained with a variety of motion-capture technologies, and including both professional and casual dancers. Using this dataset, we compare our new model against two baselines, via objective metrics and a user study, and show that both the ability to model a probability distribution, as well as being able to attend over a large motion and music context are necessary to produce interesting, diverse, and realistic dance that matches the music.

Olivier Sigaud, Ahmed Akakzia, Hugo Caselles-Dupré et al.

Autonomous discovery and direct instruction are two distinct sources of learning in children but education sciences demonstrate that mixed approaches such as assisted discovery or guided play result in improved skill acquisition. In the field of Artificial Intelligence, these extremes respectively map to autonomous agents learning from their own signals and interactive learning agents fully taught by their teachers. In between should stand teachable autotelic agents (TAA): agents that learn from both internal and teaching signals to benefit from the higher efficiency of assisted discovery. Designing such agents will enable real-world non-expert users to orient the learning trajectories of agents towards their expectations. More fundamentally, this may also be a key step to build agents with human-level intelligence. This paper presents a roadmap towards the design of teachable autonomous agents. Building on developmental psychology and education sciences, we start by identifying key features enabling assisted discovery processes in child-tutor interactions. This leads to the production of a checklist of features that future TAA will need to demonstrate. The checklist allows us to precisely pinpoint the various limitations of current reinforcement learning agents and to identify the promising first steps towards TAA. It also shows the way forward by highlighting key research directions towards the design or autonomous agents that can be taught by ordinary people via natural pedagogy.