Task-Centric Policy Optimization from Misaligned Motion Priors

Humanoid control often leverages motion priors from human demonstrations to encourage natural behaviors. However, such demonstrations are frequently suboptimal or misaligned with robotic tasks due to embodiment differences, retargeting errors, and task-irrelevant variations, causing naïve imitation to degrade task performance. Conversely, task-only reinforcement learning admits many task-optimal solutions, often resulting in unnatural or unstable motions. This exposes a fundamental limitation of linear reward mixing in adversarial imitation learning. We propose \emph{Task-Centric Motion Priors} (TCMP), a task-priority adversarial imitation framework that treats imitation as a conditional regularizer rather than a co-equal objective. TCMP maximizes task improvement while incorporating imitation signals only when they are compatible with task progress, yielding an adaptive, geometry-aware update that preserves task-feasible descent and suppresses harmful imitation under misalignment. We provide theoretical analysis of gradient conflict and task-priority stationary points, and validate our claims through humanoid control experiments demonstrating robust task performance with consistent motion style under noisy demonstrations.

Trustworthy Evaluation of Robotic Manipulation: A New Benchmark and AutoEval Methods

Driven by the rapid evolution of Vision-Action and Vision-Language-Action models, imitation learning has significantly advanced robotic manipulation capabilities. However, evaluation methodologies have lagged behind, hindering the establishment of Trustworthy Evaluation for these behaviors. Current paradigms rely on binary success rates, failing to address the critical dimensions of trust: Source Authenticity (i.e., distinguishing genuine policy behaviors from human teleoperation) and Execution Quality (e.g., smoothness and safety). To bridge these gaps, we propose a solution that combines the Eval-Actions benchmark and the AutoEval architecture. First, we construct the Eval-Actions benchmark to support trustworthiness analysis. Distinct from existing datasets restricted to successful human demonstrations, Eval-Actions integrates VA and VLA policy execution trajectories alongside human teleoperation data, explicitly including failure scenarios. This dataset is structured around three core supervision signals: Expert Grading (EG), Rank-Guided preferences (RG), and Chain-of-Thought (CoT). Building on this, we propose the AutoEval architecture: AutoEval leverages Spatio-Temporal Aggregation for semantic assessment, augmented by an auxiliary Kinematic Calibration Signal to refine motion smoothness; AutoEval Plus (AutoEval-P) incorporates the Group Relative Policy Optimization (GRPO) paradigm to enhance logical reasoning capabilities. Experiments show AutoEval achieves Spearman's Rank Correlation Coefficients (SRCC) of 0.81 and 0.84 under the EG and RG protocols, respectively. Crucially, the framework possesses robust source discrimination capabilities, distinguishing between policy-generated and teleoperated videos with 99.6% accuracy, thereby establishing a rigorous standard for trustworthy robotic evaluation. Our project and code are available at https://term-bench.github.io/.

Less Is More: Scalable Visual Navigation from Limited Data

Imitation learning provides a powerful framework for goal-conditioned visual navigation in mobile robots, enabling obstacle avoidance while respecting human preferences and social norms. However, its effectiveness depends critically on the quality and diversity of training data. In this work, we show how classical geometric planners can be leveraged to generate synthetic trajectories that complement costly human demonstrations. We train Less is More (LiMo), a transformer-based visual navigation policy that predicts goal-conditioned SE(2) trajectories from a single RGB observation, and find that augmenting limited expert demonstrations with planner-generated supervision yields substantial performance gains. Through ablations and complementary qualitative and quantitative analyses, we characterize how dataset scale and diversity affect planning performance. We demonstrate real-robot deployment and argue that robust visual navigation is enabled not by simply collecting more demonstrations, but by strategically curating diverse, high-quality datasets. Our results suggest that scalable, embodiment-specific geometric supervision is a practical path toward data-efficient visual navigation.

Towards Generalisable Imitation Learning Through Conditioned Transition Estimation and Online Behaviour Alignment

State-of-the-art imitation learning from observation methods (ILfO) have recently made significant progress, but they still have some limitations: they need action-based supervised optimisation, assume that states have a single optimal action, and tend to apply teacher actions without full consideration of the actual environment state. While the truth may be out there in observed trajectories, existing methods struggle to extract it without supervision. In this work, we propose Unsupervised Imitation Learning from Observation (UfO) that addresses all of these limitations. UfO learns a policy through a two-stage process, in which the agent first obtains an approximation of the teacher's true actions in the observed state transitions, and then refines the learned policy further by adjusting agent trajectories to closely align them with the teacher's. Experiments we conducted in five widely used environments show that UfO not only outperforms the teacher and all other ILfO methods but also displays the smallest standard deviation. This reduction in standard deviation indicates better generalisation in unseen scenarios.

Dancing in Chains: Strategic Persuasion in Academic Rebuttal via Theory of Mind

Although artificial intelligence (AI) has become deeply integrated into various stages of the research workflow and achieved remarkable advancements, academic rebuttal remains a significant and underexplored challenge. This is because rebuttal is a complex process of strategic communication under severe information asymmetry rather than a simple technical debate. Consequently, current approaches struggle as they largely imitate surface-level linguistics, missing the essential element of perspective-taking required for effective persuasion. In this paper, we introduce RebuttalAgent, the first framework to ground academic rebuttal in Theory of Mind (ToM), operationalized through a ToM-Strategy-Response (TSR) pipeline that models reviewer mental state, formulates persuasion strategy, and generates strategy-grounded response. To train our agent, we construct RebuttalBench, a large-scale dataset synthesized via a novel critique-and-refine approach. Our training process consists of two stages, beginning with a supervised fine-tuning phase to equip the agent with ToM-based analysis and strategic planning capabilities, followed by a reinforcement learning phase leveraging the self-reward mechanism for scalable self-improvement. For reliable and efficient automated evaluation, we further develop Rebuttal-RM, a specialized evaluator trained on over 100K samples of multi-source rebuttal data, which achieves scoring consistency with human preferences surpassing powerful judge GPT-4.1. Extensive experiments show RebuttalAgent significantly outperforms the base model by an average of 18.3% on automated metrics, while also outperforming advanced proprietary models across both automated and human evaluations. Disclaimer: the generated rebuttal content is for reference only to inspire authors and assist in drafting. It is not intended to replace the author's own critical analysis and response.

EquiForm: Noise-Robust SE(3)-Equivariant Policy Learning from 3D Point Clouds

Visual imitation learning with 3D point clouds has advanced robotic manipulation by providing geometry-aware, appearance-invariant observations. However, point cloud-based policies remain highly sensitive to sensor noise, pose perturbations, and occlusion-induced artifacts, which distort geometric structure and break the equivariance assumptions required for robust generalization. Existing equivariant approaches primarily encode symmetry constraints into neural architectures, but do not explicitly correct noise-induced geometric deviations or enforce equivariant consistency in learned representations. We introduce EquiForm, a noise-robust SE(3)-equivariant policy learning framework for point cloud-based manipulation. EquiForm formalizes how noise-induced geometric distortions lead to equivariance deviations in observation-to-action mappings, and introduces a geometric denoising module to restore consistent 3D structure under noisy or incomplete observations. In addition, we propose a contrastive equivariant alignment objective that enforces representation consistency under both rigid transformations and noise perturbations. Built upon these components, EquiForm forms a flexible policy learning pipeline that integrates noise-robust geometric reasoning with modern generative models. We evaluate EquiForm on 16 simulated tasks and 4 real-world manipulation tasks across diverse objects and scene layouts. Compared to state-of-the-art point cloud imitation learning methods, EquiForm achieves an average improvement of 17.2% in simulation and 28.1% in real-world experiments, demonstrating strong noise robustness and spatial generalization.

MetaWorld: Skill Transfer and Composition in a Hierarchical World Model for Grounding High-Level Instructions

Humanoid robot loco-manipulation remains constrained by the semantic-physical gap. Current methods face three limitations: Low sample efficiency in reinforcement learning, poor generalization in imitation learning, and physical inconsistency in VLMs. We propose MetaWorld, a hierarchical world model that integrates semantic planning and physical control via expert policy transfer. The framework decouples tasks into a VLM-driven semantic layer and a latent dynamics model operating in a compact state space. Our dynamic expert selection and motion prior fusion mechanism leverages a pre-trained multi-expert policy library as transferable knowledge, enabling efficient online adaptation via a two-stage framework. VLMs serve as semantic interfaces, mapping instructions to executable skills and bypassing symbol grounding. Experiments on Humanoid-Bench show MetaWorld outperforms world model-based RL in task completion and motion coherence. Our code will be found at https://anonymous.4open.science/r/metaworld-2BF4/

ConceptACT: Episode-Level Concepts for Sample-Efficient Robotic Imitation Learning

Imitation learning enables robots to acquire complex manipulation skills from human demonstrations, but current methods rely solely on low-level sensorimotor data while ignoring the rich semantic knowledge humans naturally possess about tasks. We present ConceptACT, an extension of Action Chunking with Transformers that leverages episode-level semantic concept annotations during training to improve learning efficiency. Unlike language-conditioned approaches that require semantic input at deployment, ConceptACT uses human-provided concepts (object properties, spatial relationships, task constraints) exclusively during demonstration collection, adding minimal annotation burden. We integrate concepts using a modified transformer architecture in which the final encoder layer implements concept-aware cross-attention, supervised to align with human annotations. Through experiments on two robotic manipulation tasks with logical constraints, we demonstrate that ConceptACT converges faster and achieves superior sample efficiency compared to standard ACT. Crucially, we show that architectural integration through attention mechanisms significantly outperforms naive auxiliary prediction losses or language-conditioned models. These results demonstrate that properly integrated semantic supervision provides powerful inductive biases for more efficient robot learning.

Advancing Improvisation in Human-Robot Construction Collaboration: Taxonomy and Research Roadmap

The construction industry faces productivity stagnation, skilled labor shortages, and safety concerns. While robotic automation offers solutions, construction robots struggle to adapt to unstructured, dynamic sites. Central to this is improvisation, adapting to unexpected situations through creative problem-solving, which remains predominantly human. In construction's unpredictable environments, collaborative human-robot improvisation is essential for workflow continuity. This research develops a six-level taxonomy classifying human-robot collaboration (HRC) based on improvisation capabilities. Through systematic review of 214 articles (2010-2025), we categorize construction robotics across: Manual Work (Level 0), Human-Controlled Execution (Level 1), Adaptive Manipulation (Level 2), Imitation Learning (Level 3), Human-in-Loop BIM Workflow (Level 4), Cloud-Based Knowledge Integration (Level 5), and True Collaborative Improvisation (Level 6). Analysis reveals current research concentrates at lower levels, with critical gaps in experiential learning and limited progression toward collaborative improvisation. A five-dimensional radar framework illustrates progressive evolution of Planning, Cognitive Role, Physical Execution, Learning Capability, and Improvisation, demonstrating how complementary human-robot capabilities create team performance exceeding individual contributions. The research identifies three fundamental barriers: technical limitations in grounding and dialogic reasoning, conceptual gaps between human improvisation and robotics research, and methodological challenges. We recommend future research emphasizing improved human-robot communication via Augmented/Virtual Reality interfaces, large language model integration, and cloud-based knowledge systems to advance toward true collaborative improvisation.

EvoCUA: Evolving Computer Use Agents via Learning from Scalable Synthetic Experience

The development of native computer-use agents (CUA) represents a significant leap in multimodal AI. However, their potential is currently bottlenecked by the constraints of static data scaling. Existing paradigms relying primarily on passive imitation of static datasets struggle to capture the intricate causal dynamics inherent in long-horizon computer tasks. In this work, we introduce EvoCUA, a native computer use agentic model. Unlike static imitation, EvoCUA integrates data generation and policy optimization into a self-sustaining evolutionary cycle. To mitigate data scarcity, we develop a verifiable synthesis engine that autonomously generates diverse tasks coupled with executable validators. To enable large-scale experience acquisition, we design a scalable infrastructure orchestrating tens of thousands of asynchronous sandbox rollouts. Building on these massive trajectories, we propose an iterative evolving learning strategy to efficiently internalize this experience. This mechanism dynamically regulates policy updates by identifying capability boundaries -- reinforcing successful routines while transforming failure trajectories into rich supervision through error analysis and self-correction. Empirical evaluations on the OSWorld benchmark demonstrate that EvoCUA achieves a success rate of 56.7%, establishing a new open-source state-of-the-art. Notably, EvoCUA significantly outperforms the previous best open-source model, OpenCUA-72B (45.0%), and surpasses leading closed-weights models such as UI-TARS-2 (53.1%). Crucially, our results underscore the generalizability of this approach: the evolving paradigm driven by learning from experience yields consistent performance gains across foundation models of varying scales, establishing a robust and scalable path for advancing native agent capabilities.