AI agents will be predictable in certain ways that traditional agents are not. Where and how can we leverage this predictability in order to improve social welfare? We study this question in a game-theoretic setting where one agent can pay a fixed cost to simulate the other in order to learn its mixed strategy. As a negative result, we prove that, in contrast to prior work on pure-strategy simulation, enabling mixed-strategy simulation may no longer lead to improved outcomes for both players in all so-called "generalised trust games". In fact, mixed-strategy simulation does not help in any game where the simulatee's action can depend on that of the simulator. We also show that, in general, deciding whether simulation introduces Pareto-improving Nash equilibria in a given game is NP-hard. As positive results, we establish that mixed-strategy simulation can improve social welfare if the simulator has the option to scale their level of trust, if the players face challenges with both trust and coordination, or if maintaining some level of privacy is essential for enabling cooperation.

Multi-agent reinforcement learning is a key method for training multi-robot systems over a series of episodes in which robots are rewarded or punished according to their performance; only once the system is trained to a suitable standard is it deployed in the real world. If the system is not trained enough, the task will likely not be completed and could pose a risk to the surrounding environment. Therefore, reaching high performance in a shorter training period can lead to significant reductions in time and resource consumption. We introduce Multi-Agent Reinforcement Learning guided by Language-based Inter-Robot Negotiation (MARLIN), which makes the training process both faster and more transparent. We equip robots with large language models that negotiate and debate the task, producing a plan that is used to guide the policy during training. We dynamically switch between using reinforcement learning and the negotiation-based approach throughout training. This offers an increase in training speed when compared to standard multi-agent reinforcement learning and allows the system to be deployed to physical hardware earlier. As robots negotiate in natural language, we can better understand the behaviour of the robots individually and as a collective. We compare the performance of our approach to multi-agent reinforcement learning and a large language model to show that our hybrid method trains faster at little cost to performance.

Communication is an effective mechanism for coordinating the behaviors of multiple agents, broadening their views of the environment, and to support their collaborations. In the field of multi-agent deep reinforcement learning (MADRL), agents can improve the overall learning performance and achieve their objectives by communication. Agents can communicate various types of messages, either to all agents or to specific agent groups, or conditioned on specific constraints. With the growing body of research work in MADRL with communication (Comm-MADRL), there is a lack of a systematic and structural approach to distinguish and classify existing Comm-MADRL approaches. In this paper, we survey recent works in the Comm-MADRL field and consider various aspects of communication that can play a role in designing and developing multi-agent reinforcement learning systems. With these aspects in mind, we propose 9 dimensions along which Comm-MADRL approaches can be analyzed, developed, and compared. By projecting existing works into the multi-dimensional space, we discover interesting trends. We also propose some novel directions for designing future Comm-MADRL systems through exploring possible combinations of the dimensions.

As large vision-language models (LVLMs) evolve rapidly, the demand for high-quality and diverse data to align these models becomes increasingly crucial. However, the creation of such data with human supervision proves costly and time-intensive. In this paper, we investigate the efficacy of AI feedback to scale supervision for aligning LVLMs. We introduce VLFeedback, the first large-scale vision-language feedback dataset, comprising over 82K multi-modal instructions and comprehensive rationales generated by off-the-shelf models without human annotations. To evaluate the effectiveness of AI feedback for vision-language alignment, we train Silkie, an LVLM fine-tuned via direct preference optimization on VLFeedback. Silkie showcases exceptional performance regarding helpfulness, visual faithfulness, and safety metrics. It outperforms its base model by 6.9\% and 9.5\% in perception and cognition tasks, reduces hallucination issues on MMHal-Bench, and exhibits enhanced resilience against red-teaming attacks. Furthermore, our analysis underscores the advantage of AI feedback, particularly in fostering preference diversity to deliver more comprehensive improvements. Our dataset, training code and models are available at https://vlf-silkie.github.io.

Conversational Search (CS) is the task of retrieving relevant documents from a corpus within a conversational context, combining retrieval with conversational context modeling. With the explosion of Large Language Models (LLMs), the CS field has seen major improvements with LLMs rewriting user queries, accounting for conversational context. However, engaging LLMs at inference time harms efficiency. Current methods address this by distilling embeddings from human-rewritten queries to learn the context modeling task. Yet, these approaches predominantly focus on context modeling, and only treat the contrastive component of the retrieval task within a distillation-independent loss term. To address these limitations, we propose a new distillation method, as a relaxation of the previous objective, unifying retrieval and context modeling. We relax the existing training objectives by distilling similarity scores between conversations and documents, rather than relying solely on representation learning. Our proposed distillation objective allows for more freedom in the representation space and leverages the contrastive nature of document relevance. Through experiments on Learned Sparse Retrieval (LSR) across 5 CS datasets, our approach demonstrates substantial improvements in both in-domain and out-of-domain retrieval performance, outperforming state-of-the-art with gains of up to 6 points in recall for out-of-domain datasets. Additionally, through the relaxation of the objective, we propose a multi-teacher distillation, using multiple LLMs as teachers, yielding additional gains, and outperforming the teachers themselves in in-domain experiments. Finally, analysis of the sparsity of the models reveals that our distillation allows for better control over the sparsity of the trained models.

Latent diffusion models have enabled continuous-state diffusion models to handle a variety of datasets, including categorical data. However, most methods rely on fixed pretrained embeddings, limiting the benefits of joint training with the diffusion model. While jointly learning the embedding (via reconstruction loss) and the latent diffusion model (via score matching loss) could enhance performance, our analysis shows that end-to-end training risks embedding collapse, degrading generation quality. To address this issue, we introduce CATDM, a continuous diffusion framework within the embedding space that stabilizes training. We propose a novel objective combining the joint embedding-diffusion variational lower bound with a Consistency-Matching (CM) regularizer, alongside a shifted cosine noise schedule and random dropping strategy. The CM regularizer ensures the recovery of the true data distribution. Experiments on benchmarks show that CATDM mitigates embedding collapse, yielding superior results on FFHQ, LSUN Churches, and LSUN Bedrooms. In particular, CATDM achieves an FID of 6.81 on ImageNet 256 256 with 50 steps. It outperforms non-autoregressive models in machine translation and is on a par with previous methods in text generation. These probabilistic models learn the inverse of a Markov chain that gradually converts data into pure Gaussian noise, using noise-conditioned score functions (i.e., gradients of log density), which are defined only for continuous data. The core concept is to progressively recover the original data distribution using a learned transition kernel. They offer stable and relatively efficient training procedures that contribute to their success. Recent advances, such as consistency models (Song et al., 2023; Kim et al., 2023; Luo et al., 2023), have further enhanced diffusion models by reducing the number of sampling steps, making them more practical for real-world applications.

Despite the practical success of Artificial Intelligence (AI), current neural AI algorithms face two significant issues. First, the decisions made by neural architectures are often prone to bias and brittleness. Second, when a chain of reasoning is required, neural systems often perform poorly. Neuro-symbolic artificial intelligence is a promising approach that tackles these (and other) weaknesses by combining the power of neural perception and symbolic reasoning. Meanwhile, the success of AI has made it critical to understand its behaviour, leading to the development of explainable artificial intelligence (XAI). While neuro-symbolic AI systems have important advantages over purely neural AI, we still need to explain their actions, which are obscured by the interactions of the neural and symbolic components. To address the issue, this paper proposes a formal approach to explaining the decisions of neuro-symbolic systems. The approach hinges on the use of formal abductive explanations and on solving the neuro-symbolic explainability problem hierarchically. Namely, it first computes a formal explanation for the symbolic component of the system, which serves to identify a subset of the individual parts of neural information that needs to be explained. This is followed by explaining only those individual neural inputs, independently of each other, which facilitates succinctness of hierarchical formal explanations and helps to increase the overall performance of the approach. Experimental results for a few complex reasoning tasks demonstrate practical efficiency of the proposed approach, in comparison to purely neural systems, from the perspective of explanation size, explanation time, training time, model sizes, and the quality of explanations reported.

Manipulation tasks often consist of subtasks, each representing a distinct skill. Mastering these skills is essential for robots, as it enhances their autonomy, efficiency, adaptability, and ability to work in their environment. Learning from demonstrations allows robots to rapidly acquire new skills without starting from scratch, with demonstrations typically sequencing skills to achieve tasks. Behaviour cloning approaches to learning from demonstration commonly rely on mixture density network output heads to predict robot actions. In this work, we first reinterpret the mixture density network as a library of feedback controllers (or skills) conditioned on latent states. This arises from the observation that a one-layer linear network is functionally equivalent to a classical feedback controller, with network weights corresponding to controller gains. We use this insight to derive a probabilistic graphical model that combines these elements, describing the skill acquisition process as segmentation in a latent space, where each skill policy functions as a feedback control law in this latent space. Our approach significantly improves not only task success rate, but also robustness to observation noise when trained with human demonstrations. Our physical robot experiments further show that the induced robustness improves model deployment on robots.

Advances in diffusion models for generative artificial intelligence have recently propagated to the time series (TS) domain, demonstrating state-of-the-art performance on various tasks. However, prior works on TS diffusion models often borrow the framework of existing works proposed in other domains without considering the characteristics of TS data, leading to suboptimal performance. In this work, we propose Adaptive Noise schedule for Time series diffusion models (ANT), which automatically predetermines proper noise schedules for given TS datasets based on their statistics representing non-stationarity. Our intuition is that an optimal noise schedule should satisfy the following desiderata: 1) It linearly reduces the non-stationarity of TS data so that all diffusion steps are equally meaningful, 2) the data is corrupted to the random noise at the final step, and 3) the number of steps is sufficiently large. The proposed method is practical for use in that it eliminates the necessity of finding the optimal noise schedule with a small additional cost to compute the statistics for given datasets, which can be done offline before training. We validate the effectiveness of our method across various tasks, including TS forecasting, refinement, and generation, on datasets from diverse domains. Code is available at this repository: https://github.com/seunghan96/ANT.

Current advances in Natural Language Processing (NLP) have made it increasingly feasible to build applications leveraging textual data. Generally, the core of these applications rely on having a good semantic representation of text into vectors, via embedding models. However, it has been shown that these embeddings capture and perpetuate biases already present in text. While a few techniques have been proposed to debias embeddings, they do not take advantage of the recent advances in context understanding of modern embedding models. In this paper, we fill this gap by conducting a review of 19 embedding models by quantifying their biases and how well they respond to context injection as a mean of debiasing. We show that higher performing models are more prone to capturing biases, but are also better at incorporating context. Surprisingly, we find that while models can easily embed affirmative semantics, they fail at embedding neutral semantics. Finally, in a retrieval task, we show that biases in embeddings can lead to non-desirable outcomes. We use our new-found insights to design a simple algorithm for top $k$ retrieval, where $k$ is dynamically selected. We show that our algorithm is able to retrieve all relevant gendered and neutral chunks.