Emergent communication, or emergent language, is the field of research which studies how human language-like communication systems emerge de novo in deep multi-agent reinforcement learning environments. The possibilities of replicating the emergence of a complex behavior like language have strong intuitive appeal, yet it is necessary to complement this with clear notions of how such research can be applicable to other fields of science, technology, and engineering. This paper comprehensively reviews the applications of emergent communication research across machine learning, natural language processing, linguistics, and cognitive science. Each application is illustrated with a description of its scope, an explication of emergent communication's unique role in addressing it, a summary of the extant literature working towards the application, and brief recommendations for near-term research directions.

In recent times, large language models (LLMs) have made significant strides in generating computer code, blurring the lines between code created by humans and code produced by artificial intelligence (AI). As these technologies evolve rapidly, it is crucial to explore how they influence code generation, especially given the risk of misuse in areas like higher education. This paper explores this issue by using advanced classification techniques to differentiate between code written by humans and that generated by ChatGPT, a type of LLM. We employ a new approach that combines powerful embedding features (black-box) with supervised learning algorithms - including Deep Neural Networks, Random Forests, and Extreme Gradient Boosting - to achieve this differentiation with an impressive accuracy of 98%. For the successful combinations, we also examine their model calibration, showing that some of the models are extremely well calibrated. Additionally, we present white-box features and an interpretable Bayes classifier to elucidate critical differences between the code sources, enhancing the explainability and transparency of our approach. Both approaches work well but provide at most 85-88% accuracy. We also show that untrained humans solve the same task not better than random guessing. This study is crucial in understanding and mitigating the potential risks associated with using AI in code generation, particularly in the context of higher education, software development, and competitive programming.

When answering questions, LLMs can convey not only an answer, but a level of confidence about the answer being correct. This includes explicit confidence markers (e.g. giving a numeric score) as well as implicit markers, like an authoritative tone or elaborating with additional knowledge. For LLMs to be trustworthy knowledge sources, the confidence they convey should match their actual expertise; however, most current models tend towards overconfidence. To calibrate both implicit and explicit confidence markers, we introduce a pragmatic, listener-aware finetuning method (LACIE) that models the listener, considering not only whether an answer is right, but whether it will be accepted by a listener. We cast calibration as preference optimization, creating data via a two-agent game, where a speaker model's outputs are judged by a simulated listener. We then finetune three LLMs (Mistral-7B, Llama3-8B, Llama3-70B) with LACIE, and show that the resulting models are better calibrated w.r.t. a simulated listener. Crucially, these trends transfer to human listeners, helping them correctly predict model correctness: we conduct a human evaluation where annotators accept or reject an LLM's answers, finding that training with LACIE results in 47% fewer incorrect answers being accepted while maintaining the same level of acceptance for correct answers. Furthermore, LACIE generalizes to another dataset, resulting in a large increase in truthfulness on TruthfulQA when trained on TriviaQA. Our analysis indicates that LACIE leads to a better confidence separation between correct and incorrect examples. Qualitatively, we find that a LACIE-trained model hedges more and implicitly signals certainty when it is correct by using an authoritative tone or including details. Finally, LACIE finetuning leads to an emergent increase in model abstention (e.g. saying "I don't know") for answers that are likely wrong.

Despite the success of existing instruction-tuned models, we find that they usually struggle to respond to queries with multiple instructions. This impairs their performance in complex problems whose solution consists of multiple intermediate tasks. Thus, we contend that part of the fine-tuning data mixture should be sequential--containing a chain of interrelated tasks. We first approach sequential instruction tuning from a task-driven perspective, manually creating interpretable intermediate tasks for multilingual and visual question answering: namely "translate then predict" and "caption then answer". Next, we automate this process by turning instructions in existing datasets (e.g., Alpaca and FlanCoT) into diverse and complex sequential instructions, making our method general-purpose. Models that underwent our sequential instruction tuning show improved results in coding, maths, and open-ended generation. Moreover, we put forward a new benchmark named SeqEval to evaluate a model's ability to follow all the instructions in a sequence, which further corroborates the benefits of our fine-tuning method. We hope that our endeavours will open new research avenues on instruction tuning for complex tasks.

With the emergence of the Software 3.0 era, there is a growing trend of compressing and integrating large models into software systems, with significant societal implications. Regrettably, in numerous instances, model compression techniques impact the fairness performance of these models and thus the ethical behavior of DNN-powered software. One of the most notable example is the Lottery Ticket Hypothesis (LTH), a prevailing model pruning approach. This paper demonstrates that fairness issue of LTHbased pruning arises from both its subnetwork selection and training procedures, highlighting the inadequacy of existing remedies. To address this, we propose a novel pruning framework, Ballot, which employs a novel conflict-detection-based subnetwork selection to find accurate and fair subnetworks, coupled with a refined training process to attain a high-performance model, thereby improving the fairness of DNN-powered software. By means of this procedure, Ballot improves the fairness of pruning by 38.00%, 33.91%, 17.96%, and 35.82% compared to state-of-the-art baselines, namely Magnitude Pruning, Standard LTH, SafeCompress, and FairScratch respectively, based on our evaluation of five popular datasets and three widely used models. Our code is available at https://anonymous.4open.science/r/Ballot-506E.

Heterogeneous treatment effect (HTE) estimation is vital for understanding the change of treatment effect across individuals or subgroups. Most existing HTE estimation methods focus on addressing selection bias induced by imbalanced distributions of confounders between treated and control units, but ignore distribution shifts across populations. Thereby, their applicability has been limited to the in-distribution (ID) population, which shares a similar distribution with the training dataset. In real-world applications, where population distributions are subject to continuous changes, there is an urgent need for stable HTE estimation across out-of-distribution (OOD) populations, which, however, remains an open problem. As pioneers in resolving this problem, we propose a novel Stable Balanced Representation Learning with Hierarchical-Attention Paradigm (SBRL-HAP) framework, which consists of 1) Balancing Regularizer for eliminating selection bias, 2) Independence Regularizer for addressing the distribution shift issue, 3) Hierarchical-Attention Paradigm for coordination between balance and independence. In this way, SBRL-HAP regresses counterfactual outcomes using ID data, while ensuring the resulting HTE estimation can be successfully generalized to out-of-distribution scenarios, thereby enhancing the model's applicability in real-world settings. Extensive experiments conducted on synthetic and real-world datasets demonstrate the effectiveness of our SBRL-HAP in achieving stable HTE estimation across OOD populations, with an average 10% reduction in the error metric PEHE and 11% decrease in the ATE bias, compared to the SOTA methods.

Most modern reinforcement learning algorithms optimize a cumulative single-step cost along a trajectory. The optimized motions are often 'unnatural', representing, for example, behaviors with sudden accelerations that waste energy and lack predictability. In this work, we present a novel paradigm of controlling nonlinear systems via the minimization of the Koopman spectrum cost: a cost over the Koopman operator of the controlled dynamics. This induces a broader class of dynamical behaviors that evolve over stable manifolds such as nonlinear oscillators, closed loops, and smooth movements. We demonstrate that some dynamics characterizations that are not possible with a cumulative cost are feasible in this paradigm, which generalizes the classical eigenstructure and pole assignments to nonlinear decision making. Moreover, we present a sample efficient online learning algorithm for our problem that enjoys a sub-linear regret bound under some structural assumptions.

Prompt-based interfaces for Large Language Models (LLMs) have made prototyping and building AI-powered applications easier than ever before. However, identifying potential harms that may arise from AI applications remains a challenge, particularly during prompt-based prototyping. To address this, we present Farsight, a novel in situ interactive tool that helps people identify potential harms from the AI applications they are prototyping. Based on a user's prompt, Farsight highlights news articles about relevant AI incidents and allows users to explore and edit LLM-generated use cases, stakeholders, and harms. We report design insights from a co-design study with 10 AI prototypers and findings from a user study with 42 AI prototypers. After using Farsight, AI prototypers in our user study are better able to independently identify potential harms associated with a prompt and find our tool more useful and usable than existing resources. Their qualitative feedback also highlights that Farsight encourages them to focus on end-users and think beyond immediate harms. We discuss these findings and reflect on their implications for designing AI prototyping experiences that meaningfully engage with AI harms. Farsight is publicly accessible at: https://PAIR-code.github.io/farsight.

In this dissertation I would like to guide the reader to the research on dialogue but more precisely the research I have conducted during my career since my PhD thesis. Starting from modular architectures with machine learning/deep learning and reinforcement learning to end-to-end deep neural networks. Besides my work as research associate, I also present the work I have supervised in the last years. I review briefly the state of the art and highlight the open research problems on conversational agents. Afterwards, I present my contribution to Task-Oriented Dialogues (TOD), both as research associate and as the industrial supervisor of CIFRE theses. I discuss conversational QA. Particularly, I present the work of two PhD candidates Thibault Cordier and Sebastien Montella; as well as the work of the young researcher Quentin Brabant. Finally, I present the scientific project, where I discuss about Large Language Models (LLMs) for Task-Oriented Dialogue and Multimodal Task-Oriented Dialogue.

One way to personalize and steer generations from large language models (LLM) is to assign a persona: a role that describes how the user expects the LLM to behave (e.g., a helpful assistant, a teacher, a woman). This paper investigates how personas affect diverse aspects of model behavior. We assign to seven LLMs 162 personas from 12 categories spanning variables like gender, sexual orientation, and occupation. We prompt them to answer questions from five datasets covering objective (e.g., questions about math and history) and subjective tasks (e.g., questions about beliefs and values). We also compare persona's generations to two baseline settings: a control persona setting with 30 paraphrases of "a helpful assistant" to control for models' prompt sensitivity, and an empty persona setting where no persona is assigned. We find that for all models and datasets, personas show greater variability than the control setting and that some measures of persona behavior generalize across models.