Hybrid Consistency Policy: Decoupling Multi-Modal Diversity and Real-Time Efficiency in Robotic Manipulation

In visuomotor policy learning, diffusion-based imitation learning has become widely adopted for its ability to capture diverse behaviors. However, approaches built on ordinary and stochastic denoising processes struggle to jointly achieve fast sampling and strong multi-modality. To address these challenges, we propose the Hybrid Consistency Policy (HCP). HCP runs a short stochastic prefix up to an adaptive switch time, and then applies a one-step consistency jump to produce the final action. To align this one-jump generation, HCP performs time-varying consistency distillation that combines a trajectory-consistency objective to keep neighboring predictions coherent and a denoising-matching objective to improve local fidelity. In both simulation and on a real robot, HCP with 25 SDE steps plus one jump approaches the 80-step DDPM teacher in accuracy and mode coverage while significantly reducing latency. These results show that multi-modality does not require slow inference, and a switch time decouples mode retention from speed. It yields a practical accuracy efficiency trade-off for robot policies.

Learning to Manage Investment Portfolios beyond Simple Utility Functions

While investment funds publicly disclose their objectives in broad terms, their managers optimize for complex combinations of competing goals that go beyond simple risk-return trade-offs. Traditional approaches attempt to model this through multi-objective utility functions, but face fundamental challenges in specification and parameterization. We propose a generative framework that learns latent representations of fund manager strategies without requiring explicit utility specification. Our approach directly models the conditional probability of a fund's portfolio weights, given stock characteristics, historical returns, previous weights, and a latent variable representing the fund's strategy. Unlike methods based on reinforcement learning or imitation learning, which require specified rewards or labeled expert objectives, our GAN-based architecture learns directly from the joint distribution of observed holdings and market data. We validate our framework on a dataset of 1436 U.S. equity mutual funds. The learned representations successfully capture known investment styles, such as "growth" and "value," while also revealing implicit manager objectives. For instance, we find that while many funds exhibit characteristics of Markowitz-like optimization, they do so with heterogeneous realizations for turnover, concentration, and latent factors. To analyze and interpret the end-to-end model, we develop a series of tests that explain the model, and we show that the benchmark's expert labeling are contained in our model's encoding in a linear interpretable way. Our framework provides a data-driven approach for characterizing investment strategies for applications in market simulation, strategy attribution, and regulatory oversight.

Beyond Imitation: Constraint-Aware Trajectory Generation with Flow Matching For End-to-End Autonomous Driving

Planning is a critical component of end-to-end autonomous driving. However, prevailing imitation learning methods often suffer from mode collapse, failing to produce diverse trajectory hypotheses. Meanwhile, existing generative approaches struggle to incorporate crucial safety and physical constraints directly into the generative process, necessitating an additional optimization stage to refine their outputs. To address these limitations, we propose CATG, a novel planning framework that leverages Constrained Flow Matching. Concretely, CATG explicitly models the flow matching process, which inherently mitigates mode collapse and allows for flexible guidance from various conditioning signals. Our primary contribution is the novel imposition of explicit constraints directly within the flow matching process, ensuring that the generated trajectories adhere to vital safety and kinematic rules. Secondly, CATG parameterizes driving aggressiveness as a control signal during generation, enabling precise manipulation of trajectory style. Notably, on the NavSim v2 challenge, CATG achieved 2nd place with an EPDMS score of 51.31 and was honored with the Innovation Award.

Human-in-the-loop Online Rejection Sampling for Robotic Manipulation

Reinforcement learning (RL) is widely used to produce robust robotic manipulation policies, but fine-tuning vision-language-action (VLA) models with RL can be unstable due to inaccurate value estimates and sparse supervision at intermediate steps. In contrast, imitation learning (IL) is easy to train but often underperforms due to its offline nature. In this paper, we propose Hi-ORS, a simple yet effective post-training method that utilizes rejection sampling to achieve both training stability and high robustness. Hi-ORS stabilizes value estimation by filtering out negatively rewarded samples during online fine-tuning, and adopts a reward-weighted supervised training objective to provide dense intermediate-step supervision. For systematic study, we develop an asynchronous inference-training framework that supports flexible online human-in-the-loop corrections, which serve as explicit guidance for learning error-recovery behaviors. Across three real-world tasks and two embodiments, Hi-ORS fine-tunes a pi-base policy to master contact-rich manipulation in just 1.5 hours of real-world training, outperforming RL and IL baselines by a substantial margin in both effectiveness and efficiency. Notably, the fine-tuned policy exhibits strong test-time scalability by reliably executing complex error-recovery behaviors to achieve better performance.

Supervised Reinforcement Learning: From Expert Trajectories to Step-wise Reasoning

Large Language Models (LLMs) often struggle with problems that require multi-step reasoning. For small-scale open-source models, Reinforcement Learning with Verifiable Rewards (RLVR) fails when correct solutions are rarely sampled even after many attempts, while Supervised Fine-Tuning (SFT) tends to overfit long demonstrations through rigid token-by-token imitation. To address this gap, we propose Supervised Reinforcement Learning (SRL), a framework that reformulates problem solving as generating a sequence of logical "actions". SRL trains the model to generate an internal reasoning monologue before committing to each action. It provides smoother rewards based on the similarity between the model's actions and expert actions extracted from the SFT dataset in a step-wise manner. This supervision offers richer learning signals even when all rollouts are incorrect, while encouraging flexible reasoning guided by expert demonstrations. As a result, SRL enables small models to learn challenging problems previously unlearnable by SFT or RLVR. Moreover, initializing training with SRL before refining with RLVR yields the strongest overall performance. Beyond reasoning benchmarks, SRL generalizes effectively to agentic software engineering tasks, establishing it as a robust and versatile training framework for reasoning-oriented LLMs.

ManiDP: Manipulability-Aware Diffusion Policy for Posture-Dependent Bimanual Manipulation

Recent work has demonstrated the potential of diffusion models in robot bimanual skill learning. However, existing methods ignore the learning of posture-dependent task features, which are crucial for adapting dual-arm configurations to meet specific force and velocity requirements in dexterous bimanual manipulation. To address this limitation, we propose Manipulability-Aware Diffusion Policy (ManiDP), a novel imitation learning method that not only generates plausible bimanual trajectories, but also optimizes dual-arm configurations to better satisfy posture-dependent task requirements. ManiDP achieves this by extracting bimanual manipulability from expert demonstrations and encoding the encapsulated posture features using Riemannian-based probabilistic models. These encoded posture features are then incorporated into a conditional diffusion process to guide the generation of task-compatible bimanual motion sequences. We evaluate ManiDP on six real-world bimanual tasks, where the experimental results demonstrate a 39.33$\%$ increase in average manipulation success rate and a 0.45 improvement in task compatibility compared to baseline methods. This work highlights the importance of integrating posture-relevant robotic priors into bimanual skill diffusion to enable human-like adaptability and dexterity.

PORTool: Tool-Use LLM Training with Rewarded Tree

Current tool-use large language models (LLMs) are trained on static datasets, enabling them to interact with external tools and perform multi-step, tool-integrated reasoning, which produces tool-call trajectories. However, these models imitate how a query is resolved in a generic tool-call routine, thereby failing to explore possible solutions and demonstrating limited performance in an evolved, dynamic tool-call environment. In this work, we propose PORTool, a reinforcement learning (RL) method that encourages a tool-use LLM to explore various trajectories yielding the correct answer. Specifically, this method starts with generating multiple rollouts for a given query, and some of them share the first few tool-call steps, thereby forming a tree-like structure. Next, we assign rewards to each step, based on its ability to produce a correct answer and make successful tool calls. A shared step across different trajectories receives the same reward, while different steps under the same fork receive different rewards. Finally, these step-wise rewards are used to calculate fork-relative advantages, blended with trajectory-relative advantages, to train the LLM for tool use. The experiments utilize 17 tools to address user queries, covering both time-sensitive and time-invariant topics. We conduct ablation studies to systematically justify the necessity and the design robustness of step-wise rewards. Furthermore, we compare the proposed PORTool with other training approaches and demonstrate significant improvements in final accuracy and the number of tool-call steps.

VFXMaster: Unlocking Dynamic Visual Effect Generation via In-Context Learning

Visual effects (VFX) are crucial to the expressive power of digital media, yet their creation remains a major challenge for generative AI. Prevailing methods often rely on the one-LoRA-per-effect paradigm, which is resource-intensive and fundamentally incapable of generalizing to unseen effects, thus limiting scalability and creation. To address this challenge, we introduce VFXMaster, the first unified, reference-based framework for VFX video generation. It recasts effect generation as an in-context learning task, enabling it to reproduce diverse dynamic effects from a reference video onto target content. In addition, it demonstrates remarkable generalization to unseen effect categories. Specifically, we design an in-context conditioning strategy that prompts the model with a reference example. An in-context attention mask is designed to precisely decouple and inject the essential effect attributes, allowing a single unified model to master the effect imitation without information leakage. In addition, we propose an efficient one-shot effect adaptation mechanism to boost generalization capability on tough unseen effects from a single user-provided video rapidly. Extensive experiments demonstrate that our method effectively imitates various categories of effect information and exhibits outstanding generalization to out-of-domain effects. To foster future research, we will release our code, models, and a comprehensive dataset to the community.

FastVLM: Self-Speculative Decoding for Fast Vision-Language Model Inference

Vision-language Models (VLMs) have made significant strides in visual understanding and query response generation, but often face challenges of high computational cost and inference latency due to autoregressive decoding. In this work, we introduce an imitation-learning-based Self-Speculative Decoding (SSD) framework, named FastVLM, to address these limitations. Our approach employs a lightweight draft model for token generation in an autoregressive manner, while a full model verifies these tokens non-autoregressively. Accepted tokens proceed seamlessly, while rejected tokens are corrected by the full model and used to guide the draft model's refinement. Through an imitation network, FastVLM enhances the draft model by integrating deeper level insights from the full model's architecture. Also, it maintains the performance integrity of the full model while training the draft model, achieving a balance between efficiency and accuracy. Our method speeds up the inference process by 1.55-1.85x as compared to the final layer with minimal loss in performance.

Edge Collaborative Gaussian Splatting with Integrated Rendering and Communication

Gaussian splatting (GS) struggles with degraded rendering quality on low-cost devices. To address this issue, we present edge collaborative GS (ECO-GS), where each user can switch between a local small GS model to guarantee timeliness and a remote large GS model to guarantee fidelity. However, deciding how to engage the large GS model is nontrivial, due to the interdependency between rendering requirements and resource conditions. To this end, we propose integrated rendering and communication (IRAC), which jointly optimizes collaboration status (i.e., deciding whether to engage large GS) and edge power allocation (i.e., enabling remote rendering) under communication constraints across different users by minimizing a newly-derived GS switching function. Despite the nonconvexity of the problem, we propose an efficient penalty majorization minimization (PMM) algorithm to obtain the critical point solution. Furthermore, we develop an imitation learning optimization (ILO) algorithm, which reduces the computational time by over 100x compared to PMM. Experiments demonstrate the superiority of PMM and the real-time execution capability of ILO.