We offer a novel perspective on reward modeling by formulating it as a policy discriminator, which quantifies the difference between two policies to generate a reward signal, guiding the training policy towards a target policy with desired behaviors. Based on this conceptual insight, we propose a scalable pre-training method named POLicy DiscriminAtive LeaRning (POLAR), which trains a reward model (RM) to discern identical policies and discriminate different ones. Unlike traditional reward modeling methods relying on absolute preferences, POLAR captures the relative difference between one policy and an arbitrary target policy, which is a scalable, high-level optimization objective suitable for modeling generic ranking relationships. Leveraging the POLAR pre-training paradigm, we present a series of RMs with parameter scales from 1.8B to 7B. Empirical results show that POLAR substantially outperforms traditional non-pre-trained methods, significantly enhancing RM performance. For instance, POLAR-7B could improve preference accuracy from 54.8% to 81.0% on STEM tasks and from 57.9% to 85.5% on creative writing tasks compared to SOTA baselines. POLAR also shows robust generalization capabilities in RLHF using Reinforcement Finetuning (RFT), providing reliable reward signals and markedly enhancing policy performance--improving LLaMa3.1-8B

Dexterous robotic hands are essential for performing complex manipulation tasks, yet remain difficult to train due to the challenges of demonstration collection and high-dimensional control. While reinforcement learning (RL) can alleviate the data bottleneck by generating experience in simulation, it typically relies on carefully designed, task-specific reward functions, which hinder scalability and generalization. Thus, contemporary works in dexterous manipulation have often bootstrapped from reference trajectories. These trajectories specify target hand poses that guide the exploration of RL policies and object poses that enable dense, task-agnostic rewards. However, sourcing suitable trajectories---particularly for dexterous hands---remains a significant challenge. Yet, the precise details in explicit reference trajectories are often unnecessary, as RL ultimately refines the motion.

Autonomous agents situated in real-world environments must be able to master large repertoires of skills. While a single short skill can be learned quickly, it would be impractical to learn every task independently. Instead, the agent should share knowledge across behaviors such that each task can be learned efficiently, and such that the resulting model can generalize to new tasks, especially ones that are compositions or subsets of tasks seen previously. A policy conditioned on a goal or demonstration has the potential to share knowledge between tasks if it sees enough diversity of inputs. However, these methods may not generalize to a more complex task at test time. We introduce compositional plan vectors (CPVs) to enable a policy to perform compositions of tasks without additional supervision. CPVs represent trajectories as the sum of the subtasks within them. We show that CPVs can be learned within a one-shot imitation learning framework without any additional supervision or information about task hierarchy, and enable a demonstration-conditioned policy to generalize to tasks that sequence twice as many skills as the tasks seen during training. Analogously to embeddings such as word2vec in NLP, CPVs can also support simple arithmetic operations - for example, we can add the CPVs for two different tasks to command an agent to compose both tasks, without any additional training.

Autoregressive sampling is used to create a traffic snapshot. We train a scenario generation model TrafficGen with mixed data. The detailed hyperparameters are shown in Table 4. Figure 7: Dynamics of the generated traffic scenarios. The first column is the original case. The middle columns show the generated scenarios at different timesteps.

This paper presents an approach for doing so. Given a demonstration of a path, a first network generates a path abstraction.

Proof of Lemma 2. Let U be the data set associated to ν. Proof of Lemma 3. First, we prove that the property holds for the root node. We wish to prove the property for some unexplored leaf after the iteration. This is trivial if the leaf ν is not expanded in that iteration. Suppose the leaf ν is expanded. Proof of Lemma 5. From Lemma 2, we note that Q Consider any path from the root to a leaf whose length is mK for some integer K > 0. We note that for each node ν and any of its children ν (Lemma 5).

Large Language Model (LLM)-based agents demonstrate advanced reasoning capabilities, yet practical constraints frequently limit outputs to single responses, leaving significant performance potential unrealized. This paper introduces MARINE (Multi-Agent Recursive IN-context Enhancement), a theoretically grounded framework that reconceptualizes test-time reasoning as iterative refinement of a persistent reference trajectory, fundamentally departing from conventional one-shot or multi-sample paradigms. The MARINE refinement operator systematically converts a base model's pass@N capabilities into near-optimal pass@1 performance. Rigorous theoretical analysis establishes that minimal feasible batches maximize expected performance gains under fixed invocation budgets, while logarithmically growing batch schedules ensure continuous improvement without computational constraints. Comprehensive evaluation on the BrowserComp-ZH benchmark demonstrates state-of-the-art results, with a 685B-parameter implementation achieving 46.0% pass@1 accuracy. Meanwhile, MARINE establishes a new paradigm for parameter-efficient reasoning: an 80B-parameter model augmented with MARINE matches the performance of standalone 1000B-parameter agents, reducing parameter requirements by over an order of magnitude. Notably, within a fixed computational budget, the proposed MARINE delivers higher-quality samples to alignment and optimization processes than traditional sampling-and-ranking strategies. Consequently, it has great potential to boost post-training efficiency.

Continuum manipulators offer intrinsic dexterity and safe geometric compliance for navigation within confined and obstacle-rich environments. However, their infinite-dimensional backbone deformation, unmodeled internal friction, and configuration-dependent stiffness fundamentally limit the reliability of model-based kinematic formulations, resulting in inaccurate Jacobian predictions, artificial singularities, and unstable actuation behavior. Motivated by these limitations, this work presents a complete model-less control framework that bypasses kinematic modeling by using an empirically initialized Jacobian refined online through differential convex updates. Tip motion is generated via a real-time quadratic program that computes actuator increments while enforcing tendon slack avoidance and geometric limits. A backbone-tension optimization term is introduced in this paper to regulate axial loading and suppress co-activation compression. The framework is validated across circular, pentagonal, and square trajectories, demonstrating smooth convergence, stable tension evolution, and sub-millimeter steady-state accuracy without any model calibration or parameter identification. These results establish the proposed controller as a scalable alternative to model-dependent continuum manipulation in a constrained environment.

We introduce UMI-on-Air, a framework for embodiment-aware deployment of embodiment-agnostic manipulation policies. Our approach leverages diverse, unconstrained human demonstrations collected with a handheld gripper (UMI) to train generalizable visuomotor policies. A central challenge in transferring these policies to constrained robotic embodiments-such as aerial manipulators-is the mismatch in control and robot dynamics, which often leads to out-of-distribution behaviors and poor execution. To address this, we propose Embodiment-Aware Diffusion Policy (EADP), which couples a high-level UMI policy with a low-level embodiment-specific controller at inference time. By integrating gradient feedback from the controller's tracking cost into the diffusion sampling process, our method steers trajectory generation towards dynamically feasible modes tailored to the deployment embodiment. This enables plug-and-play, embodiment-aware trajectory adaptation at test time. We validate our approach on multiple long-horizon and high-precision aerial manipulation tasks, showing improved success rates, efficiency, and robustness under disturbances compared to unguided diffusion baselines. Finally, we demonstrate deployment in previously unseen environments, using UMI demonstrations collected in the wild, highlighting a practical pathway for scaling generalizable manipulation skills across diverse-and even highly constrained-embodiments. All code, data, and checkpoints will be publicly released after acceptance. Result videos can be found at umi-on-air.github.io.