--This work aims to leverage instructional video to solve complex multi-step task-and-motion planning tasks in robotics. T owards this goal, we propose an extension of the well-established Rapidly-Exploring Random Tree (RRT) planner, which simultaneously grows multiple trees around grasp and release states extracted from the guiding video. Our key novelty lies in combining contact states and 3D object poses extracted from the guiding video with a traditional planning algorithm that allows us to solve tasks with sequential dependencies, for example, if an object needs to be placed at a specific location to be grasped later . We also investigate the generalization capabilities of our approach to go beyond the scene depicted in the instructional video. T o demonstrate the benefits of the proposed video-guided planning approach, we design a new benchmark with three challenging tasks: (i) 3D re-arrangement of multiple objects between a table and a shelf, (ii) multi-step transfer of an object through a tunnel, and (iii) transferring objects using a tray similar to a waiter transfers dishes. We demonstrate the effectiveness of our planning algorithm on several robots, including the Franka Emika Panda and the KUKA KMR iiwa . For a seamless transfer of the obtained plans to the real robot, we develop a trajectory refinement approach formulated as an optimal control problem (OCP). Traditional robot motion planning algorithms seek a collision-free path from a given starting robot configuration to a given goal robot configuration [1]. Despite the large dimensionality of the configuration space, sampling-based motion planning algorithms [2], [3] have shown to be highly effective for solving complex motion planning problems for robots, ranging from six degrees of freedom (DoFs) for industrial manipulators to tens of DoFs for humanoids [4]. Manipulation task-and-motion planning (T AMP) [5] adds an additional complexity to the problem by including movable objects in the state space. This requires the planner to discover the pick-and-place actions that connect the given start and goal robot configurations, bringing the manipulated objects from their start poses to their goal poses. INRIA, Paris This work is part of the AGIMUS project, funded by the European Union under GA no.101070165. Views and opinions expressed are, however, those of the author(s) only and do not necessarily reflect those of the European Union or the European Commission.

Continual learning for reinforcement learning agents remains a significant challenge, particularly in preserving and leveraging existing information without an external signal to indicate changes in tasks or environments. In this study, we explore the effectiveness of autoencoders in detecting new tasks and matching observed environments to previously encountered ones. Our approach integrates policy optimization with familiarity autoencoders within an end-to-end continual learning system. This system can recognize and learn new tasks or environments while preserving knowledge from earlier experiences and can selectively retrieve relevant knowledge when re-encountering a known environment. Initial results demonstrate successful continual learning without external signals to indicate task changes or reencounters, showing promise for this methodology.

Recent advances in generative AI, such as ChatGPT, have transformed access to information in education, knowledge-seeking, and everyday decision-making. However, in many developing regions, access remains a challenge due to the persistent digital divide. To help bridge this gap, we developed WaLLM - a custom AI chatbot over WhatsApp, a widely used communication platform in developing regions. Beyond answering queries, WaLLM offers several features to enhance user engagement: a daily top question, suggested follow-up questions, trending and recent queries, and a leaderboard-based reward system. Our service has been operational for over 6 months, amassing over 14.7K queries from approximately 100 users. In this paper, we present WaLLM's design and a systematic analysis of logs to understand user interactions. Our results show that 55% of user queries seek factual information. "Health and well-being" was the most popular topic (28%), including queries about nutrition and disease, suggesting users view WaLLM as a reliable source. Two-thirds of users' activity occurred within 24 hours of the daily top question. Users who accessed the "Leaderboard" interacted with WaLLM 3x as those who did not. We conclude by discussing implications for culture-based customization, user interface design, and appropriate calibration of users' trust in AI systems for developing regions.

Motivation is an important factor underlying successful learning. Previous research has demonstrated the positive effects that static interactive narrative games can have on motivation. Concurrently, advances in AI have made dynamic and adaptive approaches to interactive narrative increasingly accessible. However, limited work has explored the impact that dynamic narratives can have on learner motivation. In this paper, we compare two versions of Academical, a choice-based educational interactive narrative game about research ethics. One version employs a traditional hand-authored branching plot (i.e., static narrative) while the other dynamically sequences plots during play (i.e., dynamic narrative). Results highlight the importance of responsive content and a variety of choices for player engagement, while also illustrating the challenge of balancing pedagogical goals with the dynamic aspects of narrative. We also discuss design implications that arise from these findings. Ultimately, this work provides initial steps to illuminate the emerging potential of AI-driven dynamic narrative in educational games.

Recent advancements in large language models (LLMs) have revolutionized their ability to handle single-turn tasks, yet real-world applications demand sophisticated multi-turn interactions. This survey provides a comprehensive review of recent advancements in evaluating and enhancing multi-turn interactions in LLMs. Focusing on task-specific scenarios, from instruction following in diverse domains such as math and coding to complex conversational engagements in roleplay, healthcare, education, and even adversarial jailbreak settings, we systematically examine the challenges of maintaining context, coherence, fairness, and responsiveness over prolonged dialogues. The paper organizes current benchmarks and datasets into coherent categories that reflect the evolving landscape of multi-turn dialogue evaluation. In addition, we review a range of enhancement methodologies under multi-turn settings, including model-centric strategies (contextual learning, supervised fine-tuning, reinforcement learning, and new architectures), external integration approaches (memory-augmented, retrieval-based methods, and knowledge graph), and agent-based techniques for collaborative interactions. Finally, we discuss open challenges and propose future directions for research to further advance the robustness and effectiveness of multi-turn interactions in LLMs. Related resources and papers are available at https://github.com/yubol-cmu/Awesome-Multi-Turn-LLMs.

This thesis presents novel algorithms to advance robotic object rearrangement, a critical task for autonomous systems in applications like warehouse automation and household assistance. Addressing challenges of high-dimensional planning, complex object interactions, and computational demands, our work integrates deep learning for interaction prediction, tree search for action sequencing, and parallelized computation for efficiency. Key contributions include the Deep Interaction Prediction Network (DIPN) for accurate push motion forecasting (over 90% accuracy), its synergistic integration with Monte Carlo Tree Search (MCTS) for effective non-prehensile object retrieval (100% completion in specific challenging scenarios), and the Parallel MCTS with Batched Simulations (PMBS) framework, which achieves substantial planning speed-up while maintaining or improving solution quality. The research further explores combining diverse manipulation primitives, validated extensively through simulated and real-world experiments.

Machine learning (ML) has become a go-to solution for improving how we use, experience, and interact with technology (and the world around us). Unfortunately, studies have repeatedly shown that machine learning technologies may not provide adequate support for societal identities and experiences. Intersectionality is a sociological framework that provides a mechanism for explicitly considering complex social identities, focusing on social justice and power. While the framework of intersectionality can support the development of technologies that acknowledge and support all members of society, it has been adopted and adapted in ways that are not always true to its foundations, thereby weakening its potential for impact. To support the appropriate adoption and use of intersectionality for more equitable technological outcomes, we amplify the foundational intersectionality scholarship--Crenshaw, Combahee, and Collins (three C's), to create a socially relevant preliminary framework in developing machine-learning solutions. We use this framework to evaluate and report on the (mis)alignments of intersectionality application in machine learning literature.

We study online learning of feedforward neural networks with the sign activation function that implement functions from the unit ball in $\mathbb{R}^d$ to a finite label set $\{1, \ldots, Y\}$. First, we characterize a margin condition that is sufficient and in some cases necessary for online learnability of a neural network: Every neuron in the first hidden layer classifies all instances with some margin $γ$ bounded away from zero. Quantitatively, we prove that for any net, the optimal mistake bound is at most approximately $\mathtt{TS}(d,γ)$, which is the $(d,γ)$-totally-separable-packing number, a more restricted variation of the standard $(d,γ)$-packing number. We complement this result by constructing a net on which any learner makes $\mathtt{TS}(d,γ)$ many mistakes. We also give a quantitative lower bound of approximately $\mathtt{TS}(d,γ) \geq \max\{1/(γ\sqrt{d})^d, d\}$ when $γ\geq 1/2$, implying that for some nets and input sequences every learner will err for $\exp(d)$ many times, and that a dimension-free mistake bound is almost always impossible. To remedy this inevitable dependence on $d$, it is natural to seek additional natural restrictions to be placed on the network, so that the dependence on $d$ is removed. We study two such restrictions. The first is the multi-index model, in which the function computed by the net depends only on $k \ll d$ orthonormal directions. We prove a mistake bound of approximately $(1.5/γ)^{k + 2}$ in this model. The second is the extended margin assumption. In this setting, we assume that all neurons (in all layers) in the network classify every ingoing input from previous layer with margin $γ$ bounded away from zero. In this model, we prove a mistake bound of approximately $(\log Y)/ γ^{O(L)}$, where L is the depth of the network.

We study gradient flows for loss landscapes of fully connected feed forward neural networks with commonly used continuously differentiable activation functions such as the logistic, hyperbolic tangent, softplus or GELU function. We prove that the gradient flow either converges to a critical point or diverges to infinity while the loss converges to an asymptotic critical value. Moreover, we prove the existence of a threshold $\varepsilon>0$ such that the loss value of any gradient flow initialized at most $\varepsilon$ above the optimal level converges to it. For polynomial target functions and sufficiently big architecture and data set, we prove that the optimal loss value is zero and can only be realized asymptotically. From this setting, we deduce our main result that any gradient flow with sufficiently good initialization diverges to infinity. Our proof heavily relies on the geometry of o-minimal structures. We confirm these theoretical findings with numerical experiments and extend our investigation to real-world scenarios, where we observe an analogous behavior.

We consider the problem of online prediction for an unknown, non-explosive linear stochastic system. With a known system model, the optimal predictor is the celebrated Kalman filter. In the case of unknown systems, existing approaches based on recursive least squares and its variants may suffer from degraded performance due to the highly imbalanced nature of the regression model. This imbalance can easily lead to overfitting and thus degrade prediction accuracy. We tackle this problem by injecting an inductive bias into the regression model via {exponential forgetting}. While exponential forgetting is a common wisdom in online learning, it is typically used for re-weighting data. In contrast, our approach focuses on balancing the regression model. This achieves a better trade-off between {regression} and {regularization errors}, and simultaneously reduces the {accumulation error}. With new proof techniques, we also provide a sharper logarithmic regret bound of $O(\log^3 N)$, where $N$ is the number of observations.