Large language models are increasingly deployed as autonomous agents that must plan, act, and recover from mistakes through long-horizon interaction with environments that provide rich feedback. However, prevailing outcome-driven post-training methods (e.g., RL with verifiable rewards) primarily optimize final success signals, leaving rich environment feedback underutilized. Consequently, they often lead to distribution sharpening: the policy becomes better at reproducing a narrow set of already-successful behaviors, while failing to improve the feedback-grounded agency needed to expand problem-solving capacity (e.g., Pass@k) in long-horizon settings. To address this, we propose LEAFE (Learning Feedback-Grounded Agency from Reflective Experience), a framework that internalizes recovery agency from reflective experience. Specifically, during exploration, the agent summarizes environment feedback into actionable experience, backtracks to earlier decision points, and explores alternative branches with revised actions. We then distill these experience-guided corrections into the model through supervised fine-tuning, enabling the policy to recover more effectively in future interactions. Across a diverse set of interactive coding and agentic tasks under fixed interaction budgets, LEAFE consistently improves Pass@1 over the base model and achieves higher Pass@k than outcome-driven baselines (GRPO) and experience-based methods such as Early Experience, with gains of up to 14% on Pass@128.
Accurate probabilistic modeling of the power system restoration process is essential for resilience planning, operational decision-making, and realistic simulation of resilience events. In this work, we develop data-driven probabilistic models of the restoration process using outage data from four distribution utilities. We decompose restoration into three components: normalized restore time progression, total restoration duration, and the time to first restore. The Beta distribution provides the best-pooled fit for restore time progression, and the Uniform distribution is a defensible, parsimonious approximation for many events. Total duration is modeled as a heteroskedastic Lognormal process that scales superlinearly with event size. The time to first restore is well described by a Gamma model for moderate and large events. Together, these models provide an end-to-end stochastic model for Monte Carlo simulation, probabilistic duration forecasting, and resilience planning that moves beyond summary statistics, enabling uncertainty-aware decision support grounded in utility data.
Surgical intelligence has the potential to improve the safety and consistency of surgical care, yet most existing surgical AI frameworks remain task-specific and struggle to generalize across procedures and institutions. Although multimodal foundation models, particularly multimodal large language models, have demonstrated strong cross-task capabilities across various medical domains, their advancement in surgery remains constrained by the lack of large-scale, systematically curated multimodal data. To address this challenge, we introduce Surg$Σ$, a spectrum of large-scale multimodal data and foundation models for surgical intelligence. At the core of this framework lies Surg$Σ$-DB, a large-scale multimodal data foundation designed to support diverse surgical tasks. Surg$Σ$-DB consolidates heterogeneous surgical data sources (including open-source datasets, curated in-house clinical collections and web-source data) into a unified schema, aiming to improve label consistency and data standardization across heterogeneous datasets. Surg$Σ$-DB spans 6 clinical specialties and diverse surgical types, providing rich image- and video-level annotations across 18 practical surgical tasks covering understanding, reasoning, planning, and generation, at an unprecedented scale (over 5.98M conversations). Beyond conventional multimodal conversations, Surg$Σ$-DB incorporates hierarchical reasoning annotations, providing richer semantic cues to support deeper contextual understanding in complex surgical scenarios. We further provide empirical evidence through recently developed surgical foundation models built upon Surg$Σ$-DB, illustrating the practical benefits of large-scale multimodal annotations, unified semantic design, and structured reasoning annotations for improving cross-task generalization and interpretability.
Behavior Trees (BTs) offer a powerful paradigm for designing modular and reactive robot controllers. BT planning, an emerging field, provides theoretical guarantees for the automated generation of reliable BTs. However, BT planning typically assumes that a well-designed BT system is already grounded -- comprising high-level action models and low-level control policies -- which often requires extensive expert knowledge and manual effort. In this paper, we formalize the BT Grounding problem: the automated construction of a complete and consistent BT system. We analyze its complexity and introduce CABTO (Context-Aware Behavior Tree grOunding), the first framework to efficiently solve this challenge. CABTO leverages pre-trained Large Models (LMs) to heuristically search the space of action models and control policies, guided by contextual feedback from BT planners and environmental observations. Experiments spanning seven task sets across three distinct robotic manipulation scenarios demonstrate CABTO's effectiveness and efficiency in generating complete and consistent behavior tree systems.
We develop a conservative continuous-time stochastic control framework for treatment optimization from irregularly sampled patient trajectories. The unknown patient dynamics are modeled as a controlled stochastic differential equation with treatment as a continuous-time control. Naive model-based optimization can exploit model errors and propose out-of-support controls, so optimizing the estimated dynamics may not optimize the true dynamics. To limit extrapolation, we add a consistent signature-based MMD regularizer on path space that penalizes treatment plans whose induced trajectory distribution deviates from observed trajectories. The resulting objective minimizes a computable upper bound on the true cost. Experiments on benchmark datasets show improved robustness and performance compared to non-conservative baselines.
Recent advances in multimodal agents have improved computer-use interaction and tool-usage, yet most existing systems remain reactive, optimizing actions in isolation without reasoning about future states or long-term goals. This limits planning coherence and prevents agents from reliably solving high-level, multi-step tasks. We introduce TraceR1, a two-stage reinforcement learning framework that explicitly trains anticipatory reasoning by forecasting short-horizon trajectories before execution. The first stage performs trajectory-level reinforcement learning with rewards that enforce global consistency across predicted action sequences. The second stage applies grounded reinforcement fine-tuning, using execution feedback from frozen tool agents to refine step-level accuracy and executability. TraceR1 is evaluated across seven benchmarks, covering online computer-use, offline computer-use benchmarks, and multimodal tool-use reasoning tasks, where it achieves substantial improvements in planning stability, execution robustness, and generalization over reactive and single-stage baselines. These results show that anticipatory trajectory reasoning is a key principle for building multimodal agents that can reason, plan, and act effectively in complex real-world environments.
Benchmarking the performance of complex systems such as rail networks, renewable generation assets and national economies is central to transport planning, regulation and macroeconomic analysis. Classical frontier methods, notably Data Envelopment Analysis (DEA) and Stochastic Frontier Analysis (SFA), estimate an efficient frontier in the observed input-output space and define efficiency as distance to this frontier, but rely on restrictive assumptions on the production set and only indirectly address heterogeneity and scale effects. We propose Geometric Manifold Analysis (GeMA), a latent manifold frontier framework implemented via a productivity-manifold variational autoencoder (ProMan-VAE). Instead of specifying a frontier function in the observed space, GeMA represents the production set as the boundary of a low-dimensional manifold embedded in the joint input-output space. A split-head encoder learns latent variables that capture technological structure and operational inefficiency. Efficiency is evaluated with respect to the learned manifold, endogenous peer groups arise as clusters in latent technology space, a quotient construction supports scale-invariant benchmarking, and a local certification radius, derived from the decoder Jacobian and a Lipschitz bound, quantifies the geometric robustness of efficiency scores. We validate GeMA on synthetic data with non-convex frontiers, heterogeneous technologies and scale bias, and on four real-world case studies: global urban rail systems (COMET), British rail operators (ORR), national economies (Penn World Table) and a high-frequency wind-farm dataset. Across these domains GeMA behaves comparably to established methods when classical assumptions hold, and provides additional insight in settings with pronounced heterogeneity, non-convexity or size-related bias.
There are two prevalent model-based paradigms for combinatorial problems: 1) state-based representations, such as heuristic search, dynamic programming (DP), and decision diagrams, and 2) constraint and domain-based representations, such as constraint programming (CP), (mixed-)integer programming, and Boolean satisfiability. In this paper, we bridge the gap between the DP and CP paradigms by integrating constraint propagation into DP, enabling a DP solver to prune states and transitions using constraint propagation. To this end, we implement constraint propagation using a general-purpose CP solver in the Domain-Independent Dynamic Programming framework and evaluate using heuristic search on three combinatorial optimisation problems: Single Machine Scheduling with Time Windows, the Resource Constrained Project Scheduling Problem (RCPSP), and the Travelling Salesperson Problem with Time Windows (TSPTW). Our evaluation shows that constraint propagation significantly reduces the number of state expansions, causing our approach to solve more instances than a DP solver for Single Machine Scheduling and RCPSP, and showing similar improvements for tightly constrained TSPTW instances. The runtime performance indicates that the benefits of propagation outweigh the overhead for constrained instances, but that further work into reducing propagation overhead could improve performance further. Our work is a key step in understanding the value of constraint propagation in DP solvers, providing a model-based approach to integrating DP and CP.
Marine oil spills damage ecosystems, contaminate coastlines, and disrupt food webs, while imposing substantial economic losses on fisheries and coastal communities. Prior work has demonstrated the feasibility of containing and cleaning individual spills using a duo of autonomous surface vehicles (ASVs) equipped with a towed boom and skimmers. However, existing algorithmic approaches primarily address isolated slicks and individual ASV duos, lacking scalable methods for coordinating large robotic fleets across multiple spills representative of realistic oil-spill incidents. In this work, we propose an integrated multi-robot framework for coordinated oil-spill confinement and cleanup using autonomous ASV duos. We formulate multi-spill response as a risk-weighted minimum-latency problem, where spill-specific risk factors and service times jointly determine cumulative environmental damage. To solve this problem, we develop a hybrid optimization approach combining mixed-integer linear programming, and a tailored warm-start heuristic, enabling near-optimal routing plans for scenarios with tens of spills within minutes on commodity hardware. For physical execution, we design and analyze two tracking controllers for boom-towing ASV duos: a feedback-linearization controller with proven asymptotic stability, and a baseline PID controller. Simulation results under coupled vessel-boom dynamics demonstrate accurate path tracking for both controllers. Together, these components provide a scalable, holistic framework for rapid, risk-aware multi-robot response to large-scale oil spill disasters.
Inspection planning is concerned with computing the shortest robot path to inspect a given set of points of interest (POIs) using the robot's sensors. This problem arises in a wide range of applications from manufacturing to medical robotics. To alleviate the problem's complexity, recent methods rely on sampling-based methods to obtain a more manageable (discrete) graph inspection planning (GIP) problem. Unfortunately, GIP still remains highly difficult to solve at scale as it requires simultaneously satisfying POI-coverage and path-connectivity constraints, giving rise to a challenging optimization problem, particularly at scales encountered in real-world scenarios. In this work, we present highly scalable Mixed Integer Linear Programming (MILP) solutions for GIP that significantly advance the state-of-the-art in both runtime and solution quality. Our key insight is a reformulation of the problem's core constraints as a network flow, which enables effective MILP models and a specialized Branch-and-Cut solver that exploits the combinatorial structure of flows. We evaluate our approach on medical and infrastructure benchmarks alongside large-scale synthetic instances. Across all scenarios, our method produces substantially tighter lower bounds than existing formulations, reducing optimality gaps by 30-50% on large instances. Furthermore, our solver demonstrates unprecedented scalability: it provides non-trivial solutions for problems with up to 15,000 vertices and thousands of POIs, where prior state-of-the-art methods typically exhaust memory or fail to provide any meaningful optimality guarantees.
AI agents -- systems that plan, reason, and act using large language models -- produce non-deterministic, path-dependent behavior that cannot be fully governed at design time, where with governed we mean striking the right balance between as high as possible successful task completion rate and the legal, data-breach, reputational and other costs associated with running agents. We argue that the execution path is the central object for effective runtime governance and formalize compliance policies as deterministic functions mapping agent identity, partial path, proposed next action, and organizational state to a policy violation probability. We show that prompt-level instructions (and "system prompts"), and static access control are special cases of this framework: the former shape the distribution over paths without actually evaluating them; the latter evaluates deterministic policies that ignore the path (i.e., these can only account for a specific subset of all possible paths). In our view, runtime evaluation is the general case, and it is necessary for any path-dependent policy. We develop the formal framework for analyzing AI agent governance, present concrete policy examples (inspired by the AI act), discuss a reference implementation, and identify open problems including risk calibration and the limits of enforced compliance.
We plan to decipher the common origin of neutrino mass textures within a $Δ(27)$ based framework, in association with Type-I+II seesaw mechanism. Motivated by the diversity of texture structures used in phenomenological analyses, we examine how a single framework can reproduce them without additional assumptions. We show that the interplay between vacuum alignments, model parameters, and specific Yukawa choices naturally yields several well established neutrino mass textures in the literature. Our goal is to demonstrate that a minimal $Δ(27)$ framework can unify these viable textures within a common theoretical origin.
We investigate multiple fundamental variants of the classic coordinated motion planning (CMP) problem for unit square robots in the plane under the $L_1$ metric. In coordinated motion planning, we are given two arrangements of $k$ robots and are tasked with finding a movement schedule that minimizes a certain objective function. The two most prominent objective functions are the sum of distances traveled (Min-Sum) and the latest time of arrival (Min-Makespan). Both objectives have previously been studied extensively. We introduce a new objective function for CMP in the plane. The proposed Min-Exposure objective function defines a set of polygonal regions in the plane that provide cover and asks for a schedule with minimal elapsed time during which at least one robot is partially or fully outside of these regions. We give an $\mathcal{O}(n^4\log n)$ time algorithm that computes exposure-minimal schedules for $k=2$ robots, and an XP algorithm for arbitrary $k$. As a result of independent interest, we leverage new insights to prove that both the Min-Makespan and Min-Sum objectives are fixed-parameter tractable (FPT) parameterized by the number of robots. Our parameterized complexity results generalize known FPT results for rectangular grid graphs [Eiben, Ganian, and Kanj, SoCG'23].
This paper presents a comprehensive, spatially disaggregated dataset of Austria's natural gas and hydrogen infrastructure towards 2040. The dataset covers the complete gas transmission and distribution networks down to the medium-pressure level and integrates hydrogen expansion plans from the Austrian Gas Grid Management. Transmission infrastructure is reconstructed from ENTSOG maps, converted into a topologically consistent graph representation, and enriched with technical attributes through automated spatial matching with open-source datasets such as OpenStreetMap and Global Energy Monitor. Distribution networks and infrastructure modifications are implemented using QGas, a newly developed GIS-based tool for graph-based infrastructure manipulation. To enable forward-looking energy system analyses, the dataset explicitly represents the stage-wise transition from natural gas to hydrogen infrastructure within a single dataset. Repurposed and newly constructed hydrogen pipelines are integrated within a unified network topology using node splitting and time dependent connector elements, enabling consistent modeling of parallel natural gas and hydrogen operation over time. The resulting dataset provides a detailed representation of Austria's gas and hydrogen infrastructure, including 586 natural gas pipeline segments (5000 km), 113 repurposed segments (1250 km), and 39 newly constructed hydrogen segments (820 km), connecting 720 nodes. Moreover, it includes a comprehensive set of gas demands, biogas production facilities, storage units, electrolyzers, and compressor elements, making it directly applicable for energy system optimization models.
Robotic inspection of radioactive areas enables operators to be removed from hazardous environments; however, planning and operating in confined, cluttered environments remain challenging. These systems must autonomously reconstruct the unknown environment and cover its surfaces, whilst estimating and avoiding collisions with objects in the environment. In this paper, we propose a new planning approach based on next-best-view that enables simultaneous exploration and exploitation of the environment by reformulating the coverage path planning problem in terms of information gain. To handle obstacle avoidance under uncertainty, we extend the vector-field-inequalities framework to explicitly account for stochastic measurements of geometric primitives in the environment via chance constraints in a constrained optimal control law. The stochastic constraints were evaluated experimentally alongside the planner on a mobile manipulator in a confined environment to inspect a pipe network. These experiments demonstrate that the system can autonomously plan and execute inspection and coverage paths to reconstruct and fully cover the simplified pipe network. Moreover, the system successfully estimated geometric primitives online and avoided collisions during motion between viewpoints.
Lung cancer is a condition where there is abnormal growth of malignant cells that spread in an uncontrollable fashion in the lungs. Some common treatment strategies are surgery, chemotherapy, and radiation which aren't the best options due to the heterogeneous nature of cancer. In personalized medicine, treatments are tailored according to the individual's genetic information along with lifestyle aspects. In addition, AI-based deep learning methods can analyze large sets of data to find early signs of cancer, types of tumor, and prospects of treatment. The paper focuses on the development of personalized treatment plans using specific patient data focusing primarily on the genetic profile. Multi-Omics data from Genomics of Drug Sensitivity in Cancer have been used to build a predictive model along with machine learning techniques. The value of the target variable, LN-IC50, determines how sensitive or resistive a drug is. An XGBoost regressor is utilized to predict the drug response focusing on molecular and cellular features extracted from cancer datasets. Cross-validation and Randomized Search are performed for hyperparameter tuning to further optimize the model's predictive performance. For explanation purposes, SHAP (SHapley Additive exPlanations) was used. SHAP values measure each feature's impact on an individual prediction. Furthermore, interpreting feature relationships was performed using DeepSeek, a large language model trained to verify the biological validity of the features. Contextual explanations regarding the most important genes or pathways were provided by DeepSeek alongside the top SHAP value constituents, supporting the predictability of the model.
Remote sensing underpins crucial applications such as disaster relief and ecological field surveys, where systems must understand complex scenes and constraints and make reliable decisions. Current remote-sensing benchmarks mainly focus on evaluating perception and reasoning capabilities of multimodal large language models (MLLMs). They fail to assess planning capability, stemming either from the difficulty of curating and validating planning tasks at scale or from evaluation protocols that are inaccurate and inadequate. To address these limitations, we introduce NeSy-Route, a large-scale neuro-symbolic benchmark for constrained route planning in remote sensing. Within this benchmark, we introduce an automated data-generation framework that integrates high-fidelity semantic masks with heuristic search to produce diverse route-planning tasks with provably optimal solutions. This allows NeSy-Route to comprehensively evaluate planning across 10,821 route-planning samples, nearly 10 times larger than the largest prior benchmark. Furthermore, a three-level hierarchical neuro-symbolic evaluation protocol is developed to enable accurate assessment and support fine-grained analysis on perception, reasoning, and planning simultaneously. Our comprehensive evaluation of various state-of-the-art MLLMs demonstrates that existing MLLMs show significant deficiencies in perception and planning capabilities. We hope NeSy-Route can support further research and development of more powerful MLLMs for remote sensing.
We introduce PyPhonPlan, a Python toolkit for implementing dynamical models of phonetic planning using coupled dynamic neural fields and task dynamic simulations. The toolkit provides modular components for defining planning, perception and memory fields, as well as between-field coupling, gestural inputs, and using field activation profiles to solve tract variable trajectories. We illustrate the toolkit's capabilities through an example application:~simulating production/perception loops with a coupled memory field, which demonstrates the framework's ability to model interactive speech dynamics using representations that are temporally-principled, neurally-grounded, and phonetically-rich. PyPhonPlan is released as open-source software and contains executable examples to promote reproducibility, extensibility, and cumulative computational development for speech communication research.
Real-time LiDAR-visual-inertial odometry and mapping is crucial for navigation and planning tasks in intelligent transportation systems. This study presents a pose-only bundle adjustment (PA) LiDAR-visual-inertial odometry (LVIO), named PA-LVIO, to meet the urgent need for real-time navigation and mapping. The proposed PA framework for LiDAR and visual measurements is highly accurate and efficient, and it can derive reliable frame-to-frame constraints within multiple frames. A marginalization-free and frame-to-map (F2M) LiDAR measurement model is integrated into the state estimator to eliminate odometry drifts. Meanwhile, an IMU-centric online spatial-temporal calibration is employed to obtain a pixel-wise LiDAR-camera alignment. With accurate estimated odometry and extrinsics, a high-quality and RGB-rendered point-cloud map can be built. Comprehensive experiments are conducted on both public and private datasets collected by wheeled robot, unmanned aerial vehicle (UAV), and handheld devices with 28 sequences and more than 50 km trajectories. Sufficient results demonstrate that the proposed PA-LVIO yields superior or comparable performance to state-of-the-art LVIO methods, in terms of the odometry accuracy and mapping quality. Besides, PA-LVIO can run in real-time on both the desktop PC and the onboard ARM computer.
Motion planning under dynamics constraints, i.e., kinodynamic planning, enables safe robot operation by generating dynamically feasible trajectories that the robot can accurately track. For high-\dof robots such as manipulators, sampling-based motion planners are commonly used, especially for complex tasks in cluttered environments. However, enforcing constraints on robot dynamics in such planners requires solving either challenging two-point boundary value problems (BVPs) or propagating robot dynamics over time, both of which are computational bottlenecks that drastically increase planning times. Meanwhile, recent efforts have shown that sampling-based motion planners can generate plans in microseconds using parallelization, but are limited to geometric paths. This paper develops AkinoPDF, a fast parallelized sampling-based kinodynamic motion planning technique for a broad class of differentially flat robot systems, including manipulators, ground and aerial vehicles, and more. Differential flatness allows us to transform the motion planning problem from the original state space to a flat output space, where an analytical time-parameterized solution of the BVP and dynamics integration can be obtained. A trajectory in the flat output space is then converted back to a closed-form dynamically feasible trajectory in the original state space, enabling fast validation via ``single instruction, multiple data" parallelism. Our method is fast, exact, and compatible with any sampling-based motion planner. We extensively verify the effectiveness of our approach in both simulated benchmarks and real experiments with cluttered and dynamic environments, requiring mere microseconds to milliseconds of planning time.
Autonomous driving is undergoing a shift from modular rule based pipelines toward end to end (E2E) learning systems. This paper examines this transition by tracing the evolution from classical sense perceive plan control architectures to large driving models (LDMs) capable of mapping raw sensor input directly to driving actions. We analyze recent developments including Tesla's Full Self Driving (FSD) V12 V14, Rivian's Unified Intelligence platform, NVIDIA Cosmos, and emerging commercial robotaxi deployments, focusing on architectural design, deployment strategies, safety considerations and industry implications. A key emerging product category is supervised E2E driving, often referred to as FSD (Supervised) or L2 plus plus, which several manufacturers plan to deploy from 2026 onwards. These systems can perform most of the Dynamic Driving Task (DDT) in complex environments while requiring human supervision, shifting the driver's role to safety oversight. Early operational evidence suggests E2E learning handles the long tail distribution of real world driving scenarios and is becoming a dominant commercial strategy. We also discuss how similar architectural advances may extend beyond autonomous vehicles (AV) to other embodied AI systems, including humanoid robotics.
The COVID-19 pandemic has placed immense strain on hospital systems worldwide, leading to critical capacity challenges. This research proposes a two-part framework to optimize hospital capacity through patient relocation strategies. The first component involves developing a time series prediction model to forecast patient arrival rates. Using historical data on COVID-19 cases and hospitalizations, the model will generate accurate forecasts of future patient volumes. This will enable hospitals to proactively plan resource allocation and patient flow. The second com- ponent is a simulation model that evaluates the impact of different patient relocation strategies. The simulation will account for factors such as bed availability, staff capabilities, transportation logistics, and patient acuity to optimize the placement of patients across networked hospitals. Multiple scenarios will be tested, including inter-hospital trans- fers, use of temporary care facilities, and adaptations to discharge protocols. By combining predictive analytics and simulation modeling, this research aims to provide hospital administrators with a comprehensive decision-support tool. The proposed framework will empower them to anticipate demand, simulate relocation strategies, and imple- ment optimal policies to distribute patients and resources. Ultimately, this work seeks to enhance the resilience of healthcare systems in the face of COVID-19 and future pandemics.
We expand our novel computational method for unit commitment (UC) to include long-horizon planning. We introduce a fast novel algorithm to commit hydro-generators, provably accurately. We solve problems with thousands of generators at 5 minute market intervals. We show that our method can solve interconnect size UC problems in approximately 1 minute on a commodity hardware and that an increased planning horizon leads to sizable operational cost savings (our objective). This scale is infeasible for current state-of-the-art tools. We attain this runtime improvement by introducing a heuristic tailored for UC problems. Our method can be implemented using existing continuous optimization solvers and adapted for different applications. Combined, the two algorithms would allow an operator operating large systems with hydro units to make horizon-aware economic decisions.
Code review is a critical software engineering practice where developers review code changes before integration to ensure code quality, detect defects, and improve maintainability. In recent years, AI agents that can understand code context, plan review actions, and interact with development environments have been increasingly integrated into the code review process. However, there is limited empirical evidence to compare the effectiveness of AI agents and human reviewers in collaborative workflows. To address this gap, we conduct a large-scale empirical analysis of 278,790 code review conversations across 300 open-source GitHub projects. In our study, we aim to compare the feedback differences provided by human reviewers and AI agents. We investigate human-AI collaboration patterns in review conversations to understand how interaction shapes review outcomes. Moreover, we analyze the adoption of code suggestions provided by human reviewers and AI agents into the codebase and how adopted suggestions change code quality. We find that human reviewers provide additional feedback than AI agents, including understanding, testing, and knowledge transfer. Human reviewers exchange 11.8% more rounds when reviewing AI-generated code than human-written code. Moreover, code suggestions made by AI agents are adopted into the codebase at a significantly lower rate than suggestions proposed by human reviewers. Over half of unadopted suggestions from AI agents are either incorrect or addressed through alternative fixes by developers. When adopted, suggestions provided by AI agents produce significantly larger increases in code complexity and code size than suggestions provided by human reviewers. Our findings suggest that while AI agents can scale defect screening, human oversight remains critical for ensuring suggestion quality and providing contextual feedback that AI agents lack.
With AsgardBench we aim to evaluate visually grounded, high-level action sequence generation and interactive planning, focusing specifically on plan adaptation during execution based on visual observations rather than navigation or low-level manipulation. In the landscape of embodied AI benchmarks, AsgardBench targets the capability category of interactive planning, which is more sophisticated than offline high-level planning as it requires agents to revise plans in response to environmental feedback, yet remains distinct from low-level execution. Unlike prior embodied AI benchmarks that conflate reasoning with navigation or provide rich corrective feedback that substitutes for perception, AsgardBench restricts agent input to images, action history, and lightweight success/failure signals, isolating interactive planning in a controlled simulator without low-level control noise. The benchmark contains 108 task instances spanning 12 task types, each systematically varied through object state, placement, and scene configuration. These controlled variations create conditional branches in which a single instruction can require different action sequences depending on what the agent observes, emphasizing conditional branching and plan repair during execution. Our evaluations of leading vision language models show that performance drops sharply without visual input, revealing weaknesses in visual grounding and state tracking that ultimately undermine interactive planning. Our benchmark zeroes in on a narrower question: can a model actually use what it sees to adapt a plan when things do not go as expected?
AI-supported writing tools show strong potential for scaffolding students' learning of argumentative writing. Prior work has demonstrated the benefits of AI-supported cognitive scaffolds, such as idea exploration and argument refinement, but how these features function in authentic classroom settings remains underexplored. In this paper, we investigate the classroom integration of an AI-supported writing tool, VISAR. We deployed VISAR in an undergraduate writing course across three sections for one week each over two semesters (49 students total). Using a mixed-methods approach that combines interaction logs, writing artifact analysis, surveys, and interviews, we examine how students used VISAR features in authentic writing tasks. Our findings confirm that students appropriated AI-supported cognitive scaffolds for writing learning and achieved measurable learning gains. While prior studies suggest that students may bypass important cognitive processes when using AI writing assistants, our classroom deployment shows that when systems provide structured supports for planning and targeted generation, students naturally choose to engage with these cognition-preserving scaffolds. These learning-oriented interaction patterns were positively associated with argumentative writing quality, improved conceptual understanding, and emerging critical AI literacy, highlighting the design value of cognition-preserving features in AI writing tools. Together, these findings provide empirical evidence of how AI-supported writing scaffolds operate in authentic classroom contexts and offer design insights for future learning-oriented AI writing tools.
Autonomous driving requires safe planning, but most learning-based planners lack explicit self-correction ability: once an unsafe action is proposed, there is no mechanism to correct it. Thus, we propose CorrectionPlanner, an autoregressive planner with self-correction that models planning as motion-token generation within a propose, evaluate, and correct loop. At each planning step, the policy proposes an action, namely a motion token, and a learned collision critic predicts whether it will induce a collision within a short horizon. If the critic predicts a collision, we retain the sequence of historical unsafe motion tokens as a self-correction trace, generate the next motion token conditioned on it, and repeat this process until a safe motion token is proposed or the safety criterion is met. This self-correction trace, consisting of all unsafe motion tokens, represents the planner's correction process in motion-token space, analogous to a reasoning trace in language models. We train the planner with imitation learning followed by model-based reinforcement learning using rollouts from a pretrained world model that realistically models agents' reactive behaviors. Closed-loop evaluations show that CorrectionPlanner reduces collision rate by over 20% on Waymax and achieves state-of-the-art planning scores on nuPlan.
Simulation-to-real transfer remains a central challenge in robotics, as mismatches between simulated and real-world dynamics often lead to failures. While reinforcement learning offers a principled mechanism for adaptation, existing sim-to-real finetuning methods struggle with exploration and long-horizon credit assignment in the low-data regimes typical of real-world robotics. We introduce Simulation Distillation (SimDist), a sim-to-real framework that distills structural priors from a simulator into a latent world model and enables rapid real-world adaptation via online planning and supervised dynamics finetuning. By transferring reward and value models directly from simulation, SimDist provides dense planning signals from raw perception without requiring value learning during deployment. As a result, real-world adaptation reduces to short-horizon system identification, avoiding long-horizon credit assignment and enabling fast, stable improvement. Across precise manipulation and quadruped locomotion tasks, SimDist substantially outperforms prior methods in data efficiency, stability, and final performance. Project website and code: https://sim-dist.github.io/
Autonomous exploration in unknown environments typically relies on onboard state estimation for localisation and mapping. Existing exploration methods primarily maximise coverage efficiency, but often overlook that visual-inertial odometry (VIO) performance strongly depends on the availability of robust visual features. As a result, exploration policies can drive a robot into feature-sparse regions where tracking degrades, leading to odometry drift, corrupted maps, and mission failure. We propose a hierarchical perception-aware exploration framework for a stereo-equipped unmanned aerial vehicle (UAV) that explicitly couples exploration progress with feature observability. Our approach (i) associates each candidate frontier with an expected feature quality using a global feature map, and prioritises visually informative subgoals, and (ii) optimises a continuous yaw trajectory along the planned motion to maintain stable feature tracks. We evaluate our method in simulation across environments with varying texture levels and in real-world indoor experiments with largely textureless walls. Compared to baselines that ignore feature quality and/or do not optimise continuous yaw, our method maintains more reliable feature tracking, reduces odometry drift, and achieves on average 30\% higher coverage before the odometry error exceeds specified thresholds.
We present MiroThinker-1.7, a new research agent designed for complex long-horizon reasoning tasks. Building on this foundation, we further introduce MiroThinker-H1, which extends the agent with heavy-duty reasoning capabilities for more reliable multi-step problem solving. In particular, MiroThinker-1.7 improves the reliability of each interaction step through an agentic mid-training stage that emphasizes structured planning, contextual reasoning, and tool interaction. This enables more effective multi-step interaction and sustained reasoning across complex tasks. MiroThinker-H1 further incorporates verification directly into the reasoning process at both local and global levels. Intermediate reasoning decisions can be evaluated and refined during inference, while the overall reasoning trajectory is audited to ensure that final answers are supported by coherent chains of evidence. Across benchmarks covering open-web research, scientific reasoning, and financial analysis, MiroThinker-H1 achieves state-of-the-art performance on deep research tasks while maintaining strong results on specialized domains. We also release MiroThinker-1.7 and MiroThinker-1.7-mini as open-source models, providing competitive research-agent capabilities with significantly improved efficiency.