Tracking a target person from robot-egocentric views is crucial for developing autonomous robots that provide continuous personalized assistance or collaboration in Human-Robot Interaction (HRI) and Embodied AI. However, most existing target person tracking (TPT) benchmarks are limited to controlled laboratory environments with few distractions, clean backgrounds, and short-term occlusions. In this paper, we introduce a large-scale dataset designed for TPT in crowded and unstructured environments, demonstrated through a robot-person following task. The dataset is collected by a human pushing a sensor-equipped cart while following a target person, capturing human-like following behavior and emphasizing long-term tracking challenges, including frequent occlusions and the need for re-identification from numerous pedestrians. It includes multi-modal data streams, including odometry, 3D LiDAR, IMU, panoramic images, and RGB-D images, along with exhaustively annotated 2D bounding boxes of the target person across 48 sequences, both indoors and outdoors. Using this dataset and visual annotations, we perform extensive experiments with existing SOTA TPT methods, offering a thorough analysis of their limitations and suggesting future research directions.

Transformer-based large language models (LLMs) rely on contextual embeddings which generate different (continuous) representations for the same token depending on its surrounding context. Nonetheless, words and tokens typically have a limited number of senses (or meanings). We propose multi-sense embeddings as a drop-in replacement for each token in order to capture the range of their uses in a language. To construct a sense embedding dictionary, we apply a clustering algorithm to embeddings generated by an LLM and consider the cluster centers as representative sense embeddings. In addition, we propose a novel knowledge distillation method that leverages the sense dictionary to learn a smaller student model that mimics the senses from the much larger base LLM model, offering significant space and inference time savings, while maintaining competitive performance. Via thorough experiments on various benchmarks, we showcase the effectiveness of our sense embeddings and knowledge distillation approach. We share our code at https://github.com/Qitong-Wang/SenseDict

Eliciting information to reduce uncertainty about a latent entity is a critical task in many application domains, e.g., assessing individual student learning outcomes, diagnosing underlying diseases, or learning user preferences. Though natural language is a powerful medium for this purpose, large language models (LLMs) and existing fine-tuning algorithms lack mechanisms for strategically gathering information to refine their own understanding of the latent entity. To harness the generalization power and world knowledge of LLMs in developing effective information-gathering strategies, we propose an adaptive elicitation framework that actively reduces uncertainty on the latent entity. Since probabilistic modeling of an abstract latent entity is difficult, our framework adopts a predictive view of uncertainty, using a meta-learned language model to simulate future observations and enable scalable uncertainty quantification over complex natural language. Through autoregressive forward simulation, our model quantifies how new questions reduce epistemic uncertainty, enabling the development of sophisticated information-gathering strategies to choose the most informative next queries. In experiments on the 20 questions game, dynamic opinion polling, and adaptive student assessment, our method consistently outperforms baselines in identifying critical unknowns and improving downstream predictions, illustrating the promise of strategic information gathering in natural language settings.

We propose a general method for combinatorial online learning problems whose offline optimization problem can be solved efficiently via a dynamic programming algorithm defined by an arbitrary min-sum recurrence. Examples include online learning of Binary Search Trees, Matrix-Chain Multiplications, k -sets, Knapsacks, Rod Cuttings, and Weighted Interval Schedulings. For each of these problems we use the underlying graph of subproblems (called a multi-DAG) for defining a representation of the solutions of the dynamic programming problem by encoding them as a generalized version of paths (called multipaths). These multipaths encode each solution as a series of successive decisions or components over which the loss is linear. We then show that the dynamic programming algorithm for each problem leads to online algorithms for learning multipaths in the underlying multi-DAG. The algorithms maintain a distribution over the multipaths in a concise form as their hypothesis. More specifically we generalize the existing Expanded Hedge (Takimoto and Warmuth, 2003) and Component Hedge (Koolen et al., 2010) algorithms for the online shortest path problem to learning multipaths. Additionally, we introduce a new and faster prediction technique for Component Hedge which in our case directly samples from a distribution over multipaths, bypassing the need to decompose the distribution over multipaths into a mixture with small support.

Compositional understanding is crucial for human intelligence, yet it remains unclear whether contemporary vision models exhibit it. The dominant machine learning paradigm is built on the premise that scaling data and model sizes will improve out-of-distribution performance, including compositional generalization. We test this premise through controlled experiments that systematically vary data scale, concept diversity, and combination coverage. We find that compositional generalization is driven by data diversity, not mere data scale. Increased combinatorial coverage forces models to discover a linearly factored representational structure, where concepts decompose into additive components. We prove this structure is key to efficiency, enabling perfect generalization from few observed combinations. Evaluating pretrained models (DINO, CLIP), we find above-random yet imperfect performance, suggesting partial presence of this structure. Our work motivates stronger emphasis on constructing diverse datasets for compositional generalization, and considering the importance of representational structure that enables efficient compositional learning. Code available at https://github.com/oshapio/visual-compositional-generalization.

Domain-Incremental Learning (DIL) focuses on continual learning in non-stationary environments, requiring models to adjust to evolving domains while preserving historical knowledge. DIL faces two critical challenges in the context of imbalanced data: intra-domain class imbalance and cross-domain class distribution shifts. These challenges significantly hinder model performance, as intra-domain imbalance leads to underfitting of few-shot classes, while cross-domain shifts require maintaining well-learned many-shot classes and transferring knowledge to improve few-shot class performance in old domains. To overcome these challenges, we introduce the Dual-Balance Collaborative Experts (DCE) framework. DCE employs a frequency-aware expert group, where each expert is guided by specialized loss functions to learn features for specific frequency groups, effectively addressing intra-domain class imbalance. Subsequently, a dynamic expert selector is learned by synthesizing pseudo-features through balanced Gaussian sampling from historical class statistics. This mechanism navigates the trade-off between preserving many-shot knowledge of previous domains and leveraging new data to improve few-shot class performance in earlier tasks. Extensive experimental results on four benchmark datasets demonstrate DCE's state-of-the-art performance.

Multi-agent reinforcement learning (MARL) faces significant challenges in task sequencing and curriculum design, particularly for cooperative coordination scenarios. While curriculum learning has demonstrated success in single-agent domains, principled approaches for multi-agent coordination remain limited due to the absence of validated task complexity metrics. This approach presents a graph-based coordination complexity metric that integrates agent dependency entropy, spatial interference patterns, and goal overlap analysis to predict task difficulty in multi-agent environments. The complexity metric achieves strong empirical validation with rho = 0.952 correlation (p < 0.001) between predicted complexity and empirical difficulty determined by random agent performance evaluation. This approach evaluates the curriculum learning framework using MADDPG across two distinct coordination environments: achieving 56x performance improvement in tight coordination tasks (MultiWalker) and demonstrating systematic task progression in cooperative navigation (Simple Spread). Through systematic analysis, coordination tightness emerges as a predictor of curriculum learning effectiveness, where environments requiring strict agent interdependence benefit substantially from structured progression. This approach provides a validated complexity metric for multi-agent curriculum design and establishes empirical guidelines for multi-robot coordination applications.

Contrastive learning (CL) has recently emerged as an alternative to traditional supervised machine learning solutions by enabling rich representations from unstructured and unlabeled data. However, CL and, more broadly, self-supervised learning (SSL) methods often demand a large amount of data and computational resources, posing challenges for deployment on resource-constrained edge devices. In this work, we explore the feasibility and efficiency of SSL techniques for edge-based learning, focusing on trade-offs between model performance and energy efficiency. In particular, we analyze how different SSL techniques adapt to limited computational, data, and energy budgets, evaluating their effectiveness in learning robust representations under resource-constrained settings. Moreover, we also consider the energy costs involved in labeling data and assess how semi-supervised learning may assist in reducing the overall energy consumed to train CL models. Through extensive experiments, we demonstrate that tailored SSL strategies can achieve competitive performance while reducing resource consumption by up to 4X, underscoring their potential for energy-efficient learning at the edge.

Tutoring dialogues have gained significant attention in recent years, given the prominence of online learning and the emerging tutoring abilities of artificial intelligence (AI) agents powered by large language models (LLMs). Recent studies have shown that the strategies used by tutors can have significant effects on student outcomes, necessitating methods to predict how tutors will behave and how their actions impact students. However, few works have studied predicting tutor strategy in dialogues. Therefore, in this work we investigate the ability of modern LLMs, particularly Llama 3 and GPT-4o, to predict both future tutor moves and student outcomes in dialogues, using two math tutoring dialogue datasets. We find that even state-of-the-art LLMs struggle to predict future tutor strategy while tutor strategy is highly indicative of student outcomes, outlining a need for more powerful methods to approach this task.

--Loco-Manipulation for humanoid robots aims to enable robots to integrate mobility with upper-body tracking capabilities. Most existing approaches adopt hierarchical architectures that decompose control into isolated upper-body (manipulation) and lower-body (locomotion) policies. While this decomposition reduces training complexity, it inherently limits coordination between subsystems and contradicts the unified whole-body control exhibited by humans. We demonstrate that a single unified policy can achieve a combination of tracking accuracy, large workspace, and robustness for humanoid loco-manipulation. We propose the Unified Loco-Manipulation Controller (ULC), a single-policy framework that simultaneously tracks root velocity, root height, torso rotation, and dual-arm joint positions in an end-to-end manner, proving the feasibility of unified control without sacrificing performance. We achieve this unified control through key technologies: sequence skill acquisition for progressive learning complexity, residual action modeling for fine-grained control adjustments, command polynomial interpolation for smooth motion transitions, random delay release for robustness to deploy variations, load randomization for generalization to external disturbances, and center-of-gravity tracking for providing explicit policy gradients to maintain stability. Compared with strong baselines, ULC shows better tracking performance to disentangled methods and demonstrating larger workspace coverage. The unified dual-arm tracking enables precise manipulation under external loads while maintaining coordinated whole-body control for complex loco-manipulation tasks. The code and videos are available on our project website at https://ulc-humanoid.github.io. I. INTRODUCTION Humanoid robots, with their human-like morphology, represent a promising paradigm for versatile robotic systems capable of operating in human-designed environments.