Reasoning is the fundamental capability of large language models (LLMs). Due to the rapid progress of LLMs, there are two main issues of current benchmarks: i) these benchmarks can be crushed in a short time (less than 1 year), and ii) these benchmarks may be easily hacked. To handle these issues, we propose the ever-scalingness for building the benchmarks which are scaling over complexity against crushing, instance against hacking and exploitation, oversight for easy verification, and coverage for real-world relevance. This paper presents Nondeterministic Polynomial-time Problem Challenge (NPPC), an ever-scaling reasoning benchmark for LLMs. Specifically, the NPPC has three main modules: i) npgym, which provides a unified interface of 25 well-known NP-complete problems and can generate any number of instances with any levels of complexities, ii) npsolver, which provides a unified interface to evaluate the problem instances with both online and offline models via APIs and local deployments, respectively, and iii) npeval, which provides the comprehensive and readyto-use tools to analyze the performances of LLMs over different problems, the number of tokens, the reasoning errors and the solution errors. Extensive experiments over widely-used LLMs demonstrate: i) NPPC can successfully decrease the performances of advanced LLMs to below 10%, demonstrating that NPPC is not crushed by current models, ii) DeepSeek-R1, Claude-3.7-Sonnet, and o1/o3-mini are the most powerful LLMs, where DeepSeek-R1 can outperform Claude-3.7-Sonnet and o1/o3-mini in most NP-complete problems considered, and iii) the numbers of tokens in the advanced LLMs, e.g., Claude-3.7-Sonnet and DeepSeekR1, are observed first to increase and then decrease when the problem instances become more and more difficult. Through continuously scaling analysis, NPPC can provide critical insights into the limits of LLMs’ reasoning capabilities, exposing fundamental limitations and suggesting future directions for further improvements.

Generalizing beyond the training domain in image-based behavior cloning remains challenging. Existing methods address individual axes of generalization, workspace shifts, viewpoint changes, and cross-embodiment transfer, yet they are typically developed in isolation and often rely on complex pipelines. We introduce PALM (Perception Alignment for Local Manipulation), which leverages the invariance of local action distributions between out-of-distribution (OOD) and demonstrated domains to address these OOD shifts concurrently, without additional input modalities, model changes, or data collection. PALM modularizes the manipulation policy into coarse global components and a local policy for fine-grained actions. We reduce the discrepancy between in-domain and OOD inputs at the local policy level by enforcing local visual focus and consistent proprioceptive representation, allowing the policy to retrieve invariant local actions under OOD conditions. Experiments show that PALM limits OOD performance drops to 8% in simulation and 24% in the real world, compared to 45% and 77% for baselines.

Compared to text-driven diffusion-model-based image generation, editing is challenging due to the necessity for localized modifications. The complexity is further amplified for videos due to the additional requirement for temporal consistency. These challenges are highlighted when editing real-world videos, which, unlike synthetic videos, often feature factors like moving camera views and occlusions. In this paper, we propose RealCraft, an attention-control-based method for zero-shot real-world video editing. By swapping cross-attention in order to inject new features during the editing process and to relax the spatial-temporal attention of the edited object, we achieve localized shape-wise edits along with enhanced temporal consistency. Our model directly uses Stable Diffusion and operates without the need for additional networks or parameters. We showcase the proposed zero-shot attention-control-based method across a range of videos, demonstrating shape-wise, time-consistent and parameter-free editing in videos of up to 64 frames.

Safety-critical whole-body robot control demands reactive methods that ensure collision avoidance in real-time. Complementarity constraints and control barrier functions (CBF) have emerged as core tools for ensuring such safety constraints, and each represents a well-developed field. Despite addressing similar problems, their connection remains largely unexplored. This paper bridges this gap by formally proving the equivalence between these two methodologies for sampled-data, first-order systems, considering both single and multiple constraint scenarios. By demonstrating this equivalence, we provide a unified perspective on these techniques. This unification has theoretical and practical implications, facilitating the cross-application of robustness guarantees and algorithmic improvements between complementarity and CBF frameworks. We discuss these synergistic benefits and motivate future work in the comparison of the methods in more general cases.

The term safety in robotics is often understood as a synonym for avoidance. Although this perspective has led to progress in path planning and reactive control, a generalization of this perspective is necessary to include task semantics relevant to contact-rich manipulation tasks, especially during teleoperation and to ensure the safety of learned policies. We introduce the semantics-aware distance function and a corresponding computational method based on the Kelvin Transformation. This allows us to compute smooth distance approximations in an unbounded domain by instead solving a Laplace equation in a bounded domain. The semantics-aware distance generalizes signed distance functions by allowing the zero level set to lie inside of the object in regions where contact is allowed, effectively incorporating task semantics, such as object affordances, in an adaptive implicit representation of safe sets. In numerical experiments we show the computational viability of our method for real applications and visualize the computed function on a wrench with various semantic regions.

Uncertainties in contact dynamics and object geometry remain significant barriers to robust robotic manipulation. Caging helps mitigate these uncertainties by constraining an object's mobility without requiring precise contact modeling. Existing caging research often treats morphology and policy optimization as separate problems, overlooking their synergy. In this paper, we introduce CageCoOpt, a hierarchical framework that jointly optimizes manipulator morphology and control policy for robust caging-based manipulation. The framework employs reinforcement learning for policy optimization at the lower level and multitask Bayesian optimization for morphology optimization at the upper level. We incorporate a caging metric into both optimization levels to encourage caging configurations and thereby improve manipulation robustness. The evaluation consists of four manipulation tasks and demonstrates that co-optimizing morphology and policy improves task performance under uncertainties, establishing caging-guided co-optimization as a viable approach for robust manipulation.

An end-to-end (E2E) visuomotor policy is typically treated as a unified whole, but recent approaches using out-of-domain (OOD) data to pretrain the visual encoder have cleanly separated the visual encoder from the network, with the remainder referred to as the policy. We propose Visual Alignment Testing, an experimental framework designed to evaluate the validity of this functional separation. Our results indicate that in E2E-trained models, visual encoders actively contribute to decision-making resulting from motor data supervision, contradicting the assumed functional separation. In contrast, OOD-pretrained models, where encoders lack this capability, experience an average performance drop of 42% in our benchmark results, compared to the state-of-the-art performance achieved by E2E policies. We believe this initial exploration of visual encoders' role can provide a first step towards guiding future pretraining methods to address their decision-making ability, such as developing task-conditioned or context-aware encoders.

Federated Learning (FL) is an emerging distributed machine learning paradigm, where the collaborative training of a model involves dynamic participation of devices to achieve broad objectives. In contrast, classical machine learning (ML) typically requires data to be located on-premises for training, whereas FL leverages numerous user devices to train a shared global model without the need to share private data. Current robotic manipulation tasks are constrained by the individual capabilities and speed of robots due to limited low-latency computing resources. Consequently, the concept of cloud robotics has emerged, allowing robotic applications to harness the flexibility and reliability of computing resources, effectively alleviating their computational demands across the cloud-edge continuum. Undoubtedly, within this distributed computing context, as exemplified in cloud robotic manipulation scenarios, FL offers manifold advantages while also presenting several challenges and opportunities. In this paper, we present fundamental concepts of FL and their connection to cloud robotic manipulation. Additionally, we envision the opportunities and challenges associated with realizing efficient and reliable cloud robotic manipulation at scale through FL, where researchers adapt to design and verify FL models in either centralized or decentralized settings.

Useful robot control algorithms should not only achieve performance objectives but also adhere to hard safety constraints. Control Barrier Functions (CBFs) have been developed to provably ensure system safety through forward invariance. However, they often unnecessarily sacrifice performance for safety since they are purely reactive. Receding horizon control (RHC), on the other hand, consider planned trajectories to account for the future evolution of a system. This work provides a new perspective on safety-critical control by introducing Forward Invariance in Trajectory Spaces (FITS). We lift the problem of safe RHC into the trajectory space and describe the evolution of planned trajectories as a controlled dynamical system. Safety constraints defined over states can be converted into sets in the trajectory space which we render forward invariant via a CBF framework. We derive an efficient quadratic program (QP) to synthesize trajectories that provably satisfy safety constraints. Our experiments support that FITS improves the adherence to safety specifications without sacrificing performance over alternative CBF and NMPC methods.

We introduce the sequential multi-object robotic grasp sampling algorithm SeqGrasp that can robustly synthesize stable grasps on diverse objects using the robotic hand’s partial Degrees of Freedom (DoF). We use SeqGrasp to construct the large-scale Allegro Hand sequential grasping dataset SeqDataset and use it for training the diffusion-based sequential grasp generator SeqDiffuser. We experimentally evaluate SeqGrasp and SeqDiffuser against the state-of-the-art non-sequential multi-object grasp generation method MultiGrasp in simulation and on a real robot. The experimental results demonstrate that SeqGrasp and SeqDiffuser reach an 8.71%-43.33% higher grasp success rate than MultiGrasp. Furthermore, SeqDiffuser is approximately 1000 times faster at generating grasps than SeqGrasp and MultiGrasp. Project page: https://yulihn.github.io/SeqGrasp/.