Searcharxiv⌕ Search

arXiv subjects

Feng Zhao

Publications and source records attributed to Feng Zhao.

At least 19 recordsLinked to original sources

PRG-Fusion: Orchestrating Generative Priors with Reconstruction Evidence for Driving View Synthesis

Synthesizing photorealistic driving videos along specified trajectories is essential for scalable closed-loop simulation. However, reconstructed scenes exhibit spatially varying degradation under viewpoint shifts, while generative refinement can alter reliable content and introduce geometric inconsistencies. We propose \mbox{\textbf{PRG-Fusion}}, a framework for driving view synthesis that uses reconstruction evidence to orchestrate generative priors across regions. Our key insight is that reconstruction reliability and available support should jointly determine how previously unseen regions are generated and reliable content is preserved. We derive pixel-wise Preserve, Repair, and Generate (PRG) labels from reconstruction evidence to distinguish reliable appearance, degraded but supported content, and regions with insufficient support. At inference, these labels serve as a unified routing policy for region-aware spatiotemporal synthesis, orchestrating 3DGS appearance preservation, LiDAR-guided structural correction, and video-prior-driven content completion across Preserve, Repair, and Generate regions, respectively. We then follow a two-stage training paradigm, first establishing geometric control from sparse LiDAR projections and subsequently learning appearance control from dense 3DGS renderings. Extensive experiments on Waymo demonstrate that \mbox{\textbf{PRG-Fusion}} achieves state-of-the-art overall performance in novel trajectory video synthesis, with superior visual quality and geometric fidelity while maintaining competitive view consistency under large trajectory shifts. Our project page is available at https://pre-cici.github.io/PRG-Fusion/

cs.CV↗

Data-driven exploration of the strength-ductility synergy in refractory high-entropy alloys

Refractory high-entropy alloys (RHEAs) have attracted increasing attention for their exceptional ability to retain strength at high temperatures. However, current research mainly concentrates on compositions close to equimolar alloys, leaving a large part of the composition space unexplored. Here, we develop a multiscale framework based on a data-driven dislocation genome method that constructs energy landscapes for edge dislocation motion across various compositions. The local configurational force derived from the energy landscape is incorporated into dislocation dynamics simulations, revealing a scaling law connecting macroscale strength and atomic-scale local resistance governed by fluctuations in the local configurational force. Based on this scaling law, a new solid-solution strengthening model for edge dislocation in BCC alloys is developed, followed by a data-driven exploration of refractory alloys considering both strength and ductility. The developed method and theory are adaptable, allowing for extension to other BCC alloy systems and guiding high-throughput alloy design.

cond-mat.mtrl-sci↗

VLA-MOPD: On-Policy Distillation for Multi-Task VLA Policy Consolidation

Physical intelligence requires embodied agents to ground language in observations and execute actions while retaining useful skills. Vision-language-action (VLA) policies offer a shared interface, but task-specific post-training often produces separate specialists. Consolidation is difficult: mixed-task reinforcement learning provides sparse trajectory-level credit, while offline distillation may miss student-visited observations and endpoint distillation does not supervise intermediate action-generation states. We introduce VLA Multi-teacher On-Policy Distillation (VLA-MOPD), a protocol for consolidating flow-matching VLA specialists. After acquiring task or suite experts, a shared student collects robot rollouts and queries the instruction-routed frozen teacher at each visited observation and intermediate ordinary differential equation (ODE) state. Detached per-flow-step transition targets supervise valid action chunks through a masked local matching loss; an optional reference-policy term supports retention on other tasks. Starting from one trajectory per LIBERO task, VLA-MOPD raises mean success-once from 75.4% to 98.9%, compared with 98.15% for endpoint action distillation. It improves mean RoboTwin 2.0 success from 73.0% to 89.3%; on a Piper robot, two consolidation rounds improve average success from 70.0% to 90.0%. In a near-matched Spatial-adaptation study, reference matching limits the Object suite decrease to 3.0 points, versus 13.2 without the reference term.

cs.RO↗

ExeWorldBench: Benchmarking Code-Based Executable 3D World Modeling

Coding agents are increasingly evolving beyond reproducing what a world looks like toward constructing executable worlds that capture how a world works. This shift calls for executable world modeling, which integrates two core capabilities: world understanding, reasoning about the underlying structural, mechanical, and causal principles of a world, and executable realization, translating that understanding into a code-based system whose behavior emerges from physical interactions. However, existing benchmarks provide only fragmented evaluations centered on physical reasoning, visual fidelity, or behavioral correctness, leaving the integration of these capabilities insufficiently assessed. To bridge this gap, we introduce ExeWorldBench, to our knowledge the first benchmark to comprehensively evaluate world understanding and executable realization through code-based 3D world construction. ExeWorldBench comprises three progressively challenging tasks: Mechanism Executability, which evaluates the realization of individual mechanisms from visual observations; Compositional Executability, which extends to multi-stage systems with spatial, mechanical, and causal dependencies; and Intent-Driven Executable World Synthesis, which requires constructing executable worlds from high-level goals and physical requirements. Extensive experiments on frontier models reveal substantial limitations in executable world modeling, offering insights into the development of more physically grounded and causally coherent world construction systems.

cs.CV↗

GraphForge: Training Working Agents with Graph-Anchored Workspace Synthesis

Working agents need to read diverse files, coordinate tools, and produce deliverables. Training such agents requires tasks built on many real files with verifiable results, but few pipelines exist to synthesize this kind of data. Existing pipelines either generate files with models, which lack realism and diversity, or build tasks on real files without task-specific verifiers, leaving result quality unchecked. We introduce GraphForge, an evidence-graph based framework that grounds both the task and its verification in real files. Starting from occupation-grounded seeds for controlled diversity, GraphForge assembles a workspace of real files for each seed and builds an evidence graph over their relations. Since the task statement and rubrics are both derived from this graph, task requirements are backed by the workspace files and each criterion is anchored to the files needed to verify it. An initial rollout further tests executability, and a revision agent repairs the task and rubrics against the original files before trajectories are collected. Fine-tuning Qwen3.6-27B on 2,169 GraphForge trajectories brings GDPVal to 1445.7 (+65.7) under OpenHands, and Workspace-Bench-Lite and SpreadsheetBench II to 63.7 (+7.7) and 24.0 (+13.7) under Claude Code. Rejection fine-tuning on the SFT model's own rollouts, with candidates selected by the evidence-anchored rubrics, yields further improvements on all three benchmarks, suggesting that the rubrics provide a useful selection signal.

cs.CL↗

Scaling Agentic Data for Long-Horizon Terminal Intelligence

Terminal agents are moving beyond local operations toward completing real-world professional tasks. Short trajectories can demonstrate commands and tool calls, but they do not fully show how agents accumulate evidence through exploration, use it to inform judgments, continually adjust strategies based on feedback, and turn intermediate results into final deliverables. High-quality long-horizon trajectories capture these agentic behavioral patterns and their connections within a single problem-solving process, providing richer experience for sustained autonomous work. However, environment synthesis may complete or simplify the key work underlying these patterns in advance and thus remove the need for exploration, implementation, and verification that the agent was meant to perform. To address this, we introduce TerminalHorizon, a fully automated data synthesis engine for long-horizon terminal agents. Starting from real-world professional work, it uses independent execution to identify key work omitted during construction and uses the resulting evidence to guide environment reconstruction. To extend these patterns across stages, it progressively expands tasks under the same goal, introducing new requirements that build on prior results and call for further exploration and improvement. Using this engine, we construct 1.5K task environments and perform supervised fine-tuning of Qwen3.5-27B on the resulting dataset to obtain TerminalHorizon-27B. We achieve substantial performance gains on all 12 benchmarks spanning coding-agent tasks, agentic tasks in real-world production settings, and general reasoning. Multiple RSI case studies, each with a 12-hour research budget, further demonstrate the model's ability to continue experimenting beyond an initial solution, improve its methods based on feedback, and retain effective improvements in its final deliverables.

cs.CL↗

WorldStream: Streaming Interactive 3D Worlds with Hierarchical Planning and Persistent State

We present WorldStream, an open-ended 3D world model with three simultaneous capabilities: it extends playable terrain beyond any pre-specified map, acts on open-ended language in the running world, and restores visited places with their edits. A hierarchical planner over a typed decision model emits structured decision state: Macro fixes settlement, ecology, transport, and hydrological structure across a group of tiles; Micro specializes that contract into tile semantics under measured terrain and accepted neighbor commitments. Deterministic local code compiles accepted plans into terrain, roads, settlements, vegetation, and entities. Streaming separates their reach: plans run ahead along the predicted corridor, compilation only where movement will soon need it, geometry behind the user released. One world record retains plans, seeds, object identities and mutable state, boundary commitments, and edits by attribute, revised as the user advances and edits; returning tiles recompile from it. The same record grounds interaction: the model acts on existing entities, including stateful NPCs that update as they act, edits the present region, or attaches intent to unplanned regions. Together they yield a world expanding under bounded budgets, connected across independently planned tiles, and interactive.

cs.CV↗

REEL: Learning to Retrieve, Evaluate, and Evolve Visual Memory for Any Long-Horizon Multi-Shot Video Generation

Memory-based long-video generation aims to combine user control with long-range visual consistency. We organize this process into three functional components: a planner, an adapter, and a generator. Representative multi-shot systems couple memory to generator-specific interfaces or use predefined agent workflows, rather than independently training a memory policy for cross-shot objectives. We introduce REEL, a general-purpose visual-memory planner. Through a unified tool interface, REEL learns to retrieve references, create missing visual evidence, and update persistent memory. Backend-specific adapters translate its evidence plans into supported conditioning inputs, allowing the same policy to guide different black-box video generators. To train REEL, we construct two high-quality datasets, REEL-SFT-4k and REEL-RL-1k, covering diverse interaction patterns and prompt types. Training proceeds in three stages: supervised fine-tuning, single-shot Group Relative Policy Optimization (GRPO), and multi-shot GRPO. Multi-shot GRPO carries policy-generated memory across consecutive shots, so later outcomes guide earlier reference selection and memory updates. REEL-Bench tests identity preservation and intended state changes under text-only requests and timed reference-image uploads. On JoyAI-Echo~1.5 and MiniMax-H3, REEL improves cross-shot consistency on ST-Bench by 15.1\% and 20.7\%, and Core-100 on our proposed REEL-Bench by 2.41 and 8.32 points, respectively, over the no-memory baseline.

cs.CV↗

Adaptive Probabilistic Constellation Shaping based on Enumerative Sphere Shaping for FSO Channel with Turbulence and Pointing Errors

Free-space optical (FSO) transmission enables fast, secure, and efficient next-generation communications with abundant spectrum resources. However, atmospheric turbulence, pointing errors, path loss, and atmospheric loss induce random attenuation, challenging link reliability. Adaptive coded modulation technology enhances spectrum utilization and reliability. We propose an adaptive probabilistic constellation shaping (A-PCS) coherent system utilizing enumerative sphere shaping (ESS) for distribution matcher (DM). With PCS-64QAM, the system achieves continuous rate control from conventional QPSK-equivalent to 64QAM spectral efficiency, providing quasi-continuous control with granularities of approximately $0.05$~bits/4D for spectral efficiency and $0.1$~dB for the post-FEC SNR threshold, and a maximum control depth of $12.5$~dB. Leveraging ESS for efficient sequence utilization, it offers higher spectral efficiency and finer control granularity than constant composition distribution matcher (CCDM)-based A-PCS systems. We further model and analyze the FSO channel, presenting calculations and comparisons of outage probability and ergodic capacity under varying turbulence intensities and pointing errors. Results demonstrate 99.999~\% reliability at maximum $σ_\mathrm{R}^2 = 1.02$ and $σ_\mathrm{s} = 0.51~\mathrm{m}$, meeting requirements under severe turbulence and large pointing errors. {Furthermore, under non-ideal delayed channel state information (CSI) feedback conditions, the system adapts to varying turbulence coherence times and feedback delays, with results showing that finer modulation granularity (provided by A-PCS-ESS) enhances immunity to feedback delay, maintaining a performance advantage over conventional adaptive schemes across a range of channel environments and delay values.

eess.SP↗

A Gradient-Based Capacity Accreditation Framework in Resource Adequacy: Formulation, Computation, and Practical Implications

Probabilistic resource adequacy assessment is a cornerstone of modern capacity accreditation. This paper develops a gradient-based framework, in which capacity accreditation is interpreted as the directional derivative of a probabilistic resource adequacy metric with respect to resource capacity, that unifies two widely used accreditation approaches: Effective Load Carrying Capability (ELCC) and Marginal Reliability Impact (MRI). Under mild regularity conditions, we show that marginal ELCC and MRI yield equivalent accreditation factors, while their numerical implementations exhibit markedly different computational characteristics. Building on this framework, we demonstrate how infinitesimal perturbation analysis enables up to a $1000\times$ speedup in gradient estimation for capacity accreditation, and we implement gradient-informed search algorithms that significantly accelerate ELCC computations relative to standard bisection methods. Large-scale Monte Carlo experiments show that MRI achieves substantial runtime reductions compared to ELCC and exhibits greater robustness to perturbation step-size selection. These results provide practical guidance for implementing efficient and scalable capacity accreditation in large-scale power systems.

eess.SY↗

A compact vapor-cell optical frequency reference with fractional frequency instability around $10^{-16}$

Compact optical frequency reference with high stability is essential for field applications such as navigation and geodesy, yet vapor cell systems have remained confined to fractional instabilities over $10^{-15}$. Here, we report a molecular iodine reference that reaches an instability of $7 \times 10^{-16}$ at 1000 s and operates at the $10^{-16}$ level from 200 to 2000 s, surpassing the best reported vapor cell standards by approximately a factor of three. This achievement is enabled by a monolithic, drift immune spectroscopic unit bonded to an ultra low expansion glass substrate with precision control of key parameters.The entire system occupies only 25 L.The system achieves $5 \times 10^{-15}$ instability at 1 s and reaches the $10^{-16}$ level over the 200 to 2000 s averaging-time range, representing the first medium term stability at the $10^{-16}$ level from a compact, field ready vapor-cell reference. Our work demonstrates that $10^{-16}$ instability can be engineered into portable systems, opening a path to high precision time-keeping beyond the laboratory.

physics.atom-ph↗

AsyncTool: Evaluating the Asynchronous Function Calling Capability under Multi-Task Scenarios

Large language model (LLM)-based agents have shown strong capabilities in using external tools to solve complex tasks. However, existing evaluations often overlook the temporal dimension of tool use, especially the impact of tool response latency, and are usually limited to single-task settings. In real-world applications, multiple tasks often need to be executed concurrently, and overall efficiency depends on whether an agent can use idle time while waiting for tool responses. We refer to this capability as asynchronous tool calling. To evaluate it, we propose AsyncTool, a benchmark for assessing LLM-based agents in interactive multi-task tool-use environments with delayed tool feedback. AsyncTool presents multiple heterogeneous tasks simultaneously and simulates realistic tool response latency during execution. Using a hybrid data evolution strategy, we construct a diverse asynchronous multitasking dataset that covers multiple scenarios and tool-use patterns. We evaluate models at the step, sub-task, and task levels, and introduce efficiency-oriented metrics to measure task coordination and completion efficiency. Extensive experiments show that delayed tool feedback poses substantial challenges to current agents and leads to clear performance degradation. Models that better coordinate task switching, dependency tracking, and state maintenance achieve stronger performance on AsyncTool. Our analysis identifies key failure modes of current tool-using agents and provides practical insights for designing future systems with stronger temporal reasoning and coordination capabilities.

cs.AI↗

FACET: Preserving Source Intent and Executable State in Terminal Task Synthesis

Training terminal agents requires scalable executable supervision, yet synthesizing high-quality terminal tasks remains challenging. Each task couples an instruction, an initialized environment, a reference solution, and an executable verifier; if these artifacts are generated from inconsistent assumptions, the resulting task may be unsolvable or incorrectly evaluated. Meanwhile, multi-stage synthesis can discard the goals, dependencies, state transitions, and procedural constraints encoded in the original sources. We present FACET (Fine-grained Agentic Construction of Executable Tasks), a framework that addresses both information preservation and cross-artifact consistency. FACET reconstructs related agent skills into coherent, information-rich scenarios, then realizes and repairs the execution environment before generating the final task artifacts. The resulting container state serves as shared grounding for the instruction, solution, and verifier, while execution-based validation and targeted repair correct artifact-specific failures without unnecessarily regenerating valid components. FACET produces complex terminal tasks with dense executable checks, and successful trajectories collected from these tasks provide effective, data-efficient supervision. Fine-tuning models across multiple scales consistently improves performance on Terminal-Bench 2.1, while analyses of alternative generation schemes support the importance of environment-grounded construction for task validity and solution-verifier alignment. These results establish source-intent preservation and shared executable-state grounding as key principles for scalable terminal-task synthesis.

cs.AI↗

Alaya-EVOKE: From Linear-Scaling Supervision to Endless World

Interactive world models must support persistent memory, responsive interaction, and long-horizon generation, yet these requirements place conflicting demands on the model. Maintaining history in the denoiser context or key-value cache incurs growing cost, forcing a trade-off between session length and retained memory, while low-latency interaction relies on few-step generation whose capabilities are bounded by its teacher. Alaya-EVOKE (Evoke) addresses both limitations by externalizing persistent world state and redesigning the teacher for long-horizon interactive generation. Scene geometry is maintained in an external, camera-indexed world state bank, from which only view-relevant information is retrieved, keeping the denoiser context bounded as the session grows. Rather than treating the teacher as a fixed generator, we design it for long-horizon supervision: its sparse attention combines chunk-wise grouping, retrieval of selected distant frames, and a linear-attention global state, yielding linear growth in memory and compute while enabling supervision over long horizons. Such supervision exposes content drift that stays locally plausible within short windows, while per-chunk conditioning enables prompt changes and event control throughout the sequence. A 30-second distribution-matching objective, applied under self-forced rollouts, transfers both capabilities to a three-step student that uses no classifier-free guidance, improving resistance to long-term drift while preserving responsive conditioning. With bounded context and recurrent external memory, Evoke supports open-ended, continuously evolving generation; on a single H200 at $384\times 640$, each $1.5\,\mathrm{s}$ chunk is generated in $2.11\,\mathrm{s}$. As a three-step world model, Evoke achieves state-of-the-art performance on WBench while remaining competitive on VBench-Long and VBench-2.0.

cs.CV↗

SynBoost: A Synergistic Framework for Fast Sampling of Diffusion Models

Diffusion probabilistic models (DPMs) have demonstrated remarkable success in visual generation. However, their iterative sampling mechanism results in slow inference speeds. While reducing sampling steps offers an intuitive acceleration strategy, it introduces significant discretization error. Existing fast samplers have made substantial progress in mitigating this error through high-order solvers, yet further optimization appears constrained. This limitation prompts a critical question: can sampling efficiency be advanced beyond current paradigms? In this work, we re-examine the composition of sampling errors and identify two distinct components: the well-studied discretization error and the under-explored approximation error. By implementing a dual-error disentanglement strategy, we elucidate the dynamic interplay between these error types across sampling steps. Building on this empirical analysis, we propose SynBoost, a unified and training-free acceleration framework that simultaneously addresses both error sources to minimize total sampling error. Concretely, it mitigates approximation error by blending the current noise estimation with a more accurate prediction from a larger timestep. SynBoost seamlessly integrates with existing samplers, substantially enhancing their speed and output quality, particularly in regimes with extremely few steps. We validate our framework through extensive experiments across unconditional and conditional generation tasks, encompassing both pixel-space and latent-space DPMs.

cs.CV↗

VideoCoCo: Code-as-CoT for Physically-Consistent Video Generation via an Agentic Dual-Engine System

Text-to-video models have achieved remarkable visual quality, yet they still struggle to generate physically consistent dynamics because the temporal evolution of a scene must be inferred implicitly from a highly compressed text prompt. Existing chain-of-thought approaches introduce intermediate plans or visual states, but these representations are typically non-executable or temporally sparse, limiting their ability to instantiate and control the complete spatiotemporal process. To address this limitation, we introduce VideoCoCo, an agentic dual-engine framework in which executable Blender code serves as a process-level chain of thought. Given a text prompt, a coding agent synthesizes a Blender program that explicitly specifies the scene and its temporal evolution. The executable simulation engine runs the program to produce a deterministic spatiotemporal draft, which is subsequently transformed into a photorealistic video by a generative video engine through draft-conditioned editing. This decomposition separates process-level reasoning from high-fidelity visual realization. To adapt the video editor to simulated drafts, we construct VideoCoCo-3K, a curated dataset of draft-instruction-target triplets. VideoCoCo improves the OmniWeaving baseline from 0.475 to 0.558 on PhyGenBench and from 52.18 to 77.88 on VBench-2.0, achieving the best average score on both benchmarks. These results demonstrate that executable code provides an effective, controllable, and inspectable intermediate representation for physically consistent video generation.

cs.CV↗

Quantifying the Availability of Synchronized and Non-Synchronized Generating Units When Needed

Do synchronized units have higher probabilities of being available when needed than non-synchronized units? Power system operation implicitly relies on the qualitative belief that synchronized units are more likely to be available when needed because they are already synchronized to the grid, whereas non-synchronized units must first start and synchronize before becoming available. However, this distinction is rarely expressed through an explicit quantitative measure. To quantify this distinction, we propose failure probabilities for synchronized and non-synchronized generating units, denoted by SynFORd and NonSynFORd, by accounting for their different initial operating states. The complements of the proposed probabilities directly represent the corresponding availability probabilities when the units are needed. Closed-form analytical expressions are derived, revealing the dominant failure mechanisms of the two unit types. Case studies using generating-unit data from the New England system show that non-synchronized units generally exhibit higher and more dispersed failure probabilities than synchronized units.

eess.SY↗

Video-DeepResearch: Towards the Next-Generation Multimodal Deepresearch Agent

We introduce Video-DeepResearch (Video-DR), extending multimodal agents from static images to continuous video streams, a setting that demands dense spatiotemporal grounding coupled with open-web exploration. Preliminary evaluations reveal two critical bottlenecks in current models: (1) modality bias, where agents bypass visual tools in favor of textual search, and (2) parametric knowledge leakage, where models rely on internal memory rather than genuine tool-augmented execution. To address these challenges, we propose Video-DR, featuring a decoupled perception-exploration pipeline with stage-wise tool unlocking that compels exhaustive cross-frame visual grounding prior to web retrieval. Our framework adopts a two-stage training recipe: supervised fine-tuning followed by Group Relative Policy Optimization (GRPO), enabling autonomous exploration that breaks the imitation-learning ceiling. Furthermore, we curate Video-DR-Bench, a human-AI collaborative benchmark comprising 200 complex, multi-hop VQA instances. Empirical results demonstrate that our Video-DeepResearch-35B-A3B establishes a new state-of-the-art of 64.0% average accuracy, surpassing proprietary Claude-4.5-Sonnet (59.0%) by 5.0 points and significantly outperforming GPT-5 (52.5%) and Gemini 2.5 Pro (57.5%). The 30B-A3B variant achieves 59.3%, competitive with Claude-4.5-Sonnet and demonstrating the effectiveness of our training paradigm even at compact scale. Code: https://github.com/Osilly/Vision-DeepResearch.

cs.CV↗