arXiv subjects

Jiaqi Liu

Publications and source records attributed to Jiaqi Liu.

At least 37 records · Page 2Linked to original sources

arXiv2026

CUA-Sandbox: Efficient Environments for Computer-Use Agent Reinforcement Learning

Reinforcement learning enables computer-use agents to improve through interaction with real software environments, including websites and desktop applications. However, conventional deployments replicate an initialized runtime for each independent rollout, even when trajectories use the same software, incurring repeated memory and initialization costs as the number of parallel environments grows. Does an independent computer-use environment require an independent execution runtime? Our key observation is that trajectories require independent mutable state, while initialized application runtimes can be reused across concurrently evolving environments, making state the natural unit of environment independence. Guided by this observation, we introduce CUA-Sandbox, which separates private state capsules from shared runtimes through state-scoped execution and transactional lifecycle operations, including resets and branches, while retaining the original software interfaces and task evaluators. Experiments show comparable or improved task success relative to Docker, while substantially reducing rollout and resource costs. CUA-Sandbox achieves up to a 6.20x increase in rollout throughput, a 9.2x reduction in per-environment memory, and a 504x reduction in incremental storage.

cs.AI↗

arXiv2026

DriveHierarchy: A Benchmark for Diagnosing VLM Driving Capabilities from Open-Loop Understanding to Closed-Loop Execution

Evaluating VLM-based autonomous driving remains difficult because driving competence is composite, where a capable system must ground traffic participants and hazards, integrate context across views and time, reason about future evolution, and act appropriately under closed-loop interaction. Existing benchmarks usually assess either open-loop understanding or closed-loop driving but provide limited structure for explaining how these abilities are organized, how they relate, and how they may inform model diagnosis and improvement. We present \textsc{DriveHierarchy}, a hierarchical benchmark that organizes VLM-based autonomous driving into four ranks, spanning perceptual grounding, contextual memory, mental reasoning, and closed-loop execution. To instantiate this hierarchy, we integrate multiple open-source autonomous-driving datasets into a unified open-loop benchmark with 76,798 question-answer pairs over 84,279 frames and develop a closed-loop simulation platform with interactive scenario construction on a real-world road network, from which 100 driving scenarios are curated for embodied evaluation. Experiments on 15 VLMs show that \textsc{DriveHierarchy} captures structured but non-redundant capability variation, relates open-loop understanding to closed-loop driving, and provides a practical basis for diagnosis and benchmark-guided optimization. \textsc{DriveHierarchy} therefore serves as a unified framework for evaluating and improving VLM-based autonomous driving systems. An anonymized project has been released on https://github.com/PerfectXu88/DriveHierarchy

cs.AI↗

arXiv2026

Random dynamical systems generated by rough differential equations driven by controlled rough paths

We study rough differential equations driven by controlled rough paths with drift. By means of the canonical lift of the controlled driver, we establish pathwise well-posedness and continuity. Under compatible time-shift conditions, we construct differentiable solution cocycles for the associated dynamical system generated by the equations considered here. Furthermore, assuming the corresponding regularity, moment, and spectral hypotheses, we prove the invariance of the local stable, unstable, and center manifolds.

math.PR↗

arXiv2026

Agent0: Unleashing Self-Evolving Agents from Zero Data via Tool-Integrated Reasoning

Large Language Model (LLM) Agents, often trained with Reinforcement Learning (RL), are constrained by a dependency on human-curated data, limiting scalability and tethering AI to human knowledge. Existing self-evolution frameworks offer an alternative but are typically restricted by the model's inherent capabilities and single-round interactions, hindering the development of complex curricula involving tool use or dynamic reasoning. We introduce Agent0, a fully autonomous framework that evolves high-performing agents without external data through multi-step co-evolution and seamless tool integration. Agent0 establishes a symbiotic competition between two agents initialized from the same base LLM: a curriculum agent that proposes increasingly challenging frontier tasks, and an executor agent that learns to solve them. We integrate external tools to enhance the executor's problem-solving capacity; this improvement, in turn, pressures the curriculum agent to construct more complex, tool-aware tasks. Through this iterative process, Agent0 establishes a self-reinforcing cycle that continuously produces high-quality curricula. Empirically, Agent0 substantially boosts reasoning capabilities, improving the Qwen3-8B-Base model by 18% on mathematical reasoning and 24% on general reasoning benchmarks. Code is available at https://github.com/aiming-lab/Agent0.

cs.LG↗

arXiv2026

MCPGen: Benchmarking LLMs on Executable MCPWorkflow Development

We study whether LLMs can produce executable workflow artifacts that remain consistent across graph structure, tool implementation, schema bindings, and runtime wiring. In this setting, correctness depends on cross-layer consistency: a workflow may be structurally plausible, yet still fail because tool implementations, schema bindings, or runtime execution do not align. Existing benchmarks largely evaluate these capabilities in isolation or rely on trajectory-level proxies, leaving open whether generated workflow artifacts execute end-to-end. We introduce \textbf{MCPGen}, an executable benchmark for Model Context Protocol (MCP) workflow development. MCPGen contains 100 self-contained MCP projects across 16 application domains and evaluates three diagnostic tasks: workflow reconstruction, tool creation, and backward-compatible workflow extension. We evaluate 11 representative LLMs in a single-turn foundation-model setting, assessing generated artifacts through static analysis, unit and integration tests, and process-isolated end-to-end execution. Models reach 88.5\% on workflow reconstruction, but no model exceeds 57\% end-to-end execution success. Per-tool unit-test pass rates reach 63.8\%, while project-level integration success does not exceed 45\%, suggesting that integration remains a major bottleneck even when isolated tool tests pass.

cs.SE↗

arXiv2026

CVP Is NP-Complete for Principal Cyclotomic Ideals

We prove that exact Euclidean decision-CVP is $\mathsf{NP}$-complete on the coefficient lattices of nonzero principal ideals in the power-of-two cyclotomic rings $R_d:=\mathbb{Z}[y]/(y^d+1)$. Our deterministic reduction from Exact Cover by 3-Sets (X3C) produces a target and a squared threshold $Δ$ such that the closest squared distance is exactly $Δ$ in YES instances and at least $Δ+4$ in NO instances. This also implies $\mathsf{NP}$-hardness of exact search-CVP under polynomial-time Turing reductions. We also transfer the resulting principal-ideal CVP instances to full-rank principal ideals of the cyclic quotient ring $\mathbb{Z}[X]/(X^D-1)$, where $D:=2d$. Their coefficient lattices are invariant under cyclic coordinate shifts. The lift preserves principality and multiplies corresponding squared distances by eight. Thus, on principal cyclic ideal lattices, exact decision-CVP is $\mathsf{NP}$-complete and exact search-CVP is $\mathsf{NP}$-hard. We also obtain uniformly computable fixed cyclotomic and cyclic families in which only the target and threshold depend on the X3C collection. Consequently, a polynomial-time solution to exact decision-CVPP on either family would imply $\mathsf{NP}\subseteq\mathsf{P}/\mathrm{poly}$ and collapse the polynomial hierarchy to $Σ_2^{\mathsf{P}}$. To our knowledge, the cyclic results answer Micciancio's questions of whether exact decision-CVP is $\mathsf{NP}$-hard on cyclic lattices and on a fixed family of cyclic lattices, even when restricted to full-rank principal cyclic ideals. Finally, under the coefficient embedding, we prove that exact decision-module-SIVP is $\mathsf{NP}$-complete on free rank-two modules over the same cyclotomic rings.

cs.CC↗

arXiv2026

Faster SVP in Polynomial Space

Kannan's algorithm, as analyzed by Hanrot and Stehlé in 2007, solves the exact Euclidean shortest vector problem in polynomial space and $n^{\frac{n}{2e}+o(n)}$ time. In the classical setting with polynomial space, we obtain the first improvement on this bound via a randomized algorithm that runs in $n^{\frac{n}{4e}+o(n)}$ time. The main idea is to represent a fixed shortest vector in many ways as a difference of samples, thereby enabling the low-space collision search of Lyu and Zhu (SODA 2023) to replace exhaustive enumeration in the original analysis.

cs.DS↗

arXiv2026

TorchCraft: Unified binder design by inverting an all-atom structure predictor

All-atom structure predictors model diverse molecular interactions, but using their learned structural priors for binder design remains challenging. Here we present TorchCraft, a unified binder-design framework that optimizes sequence logits through a frozen all-atom predictor. Implemented in TorchFold, TorchCraft combines confidence, contact, geometric, and sequence-prior objectives within a shared optimization procedure for minibinders, framework-conditioned VHHs, cyclic peptides, and ligand-binding proteins. Using pretrained AlphaFold 3 weights, TorchCraft generated representative minibinders and VHHs with experimentally measured binding across four targets in each format, without post hoc sequence redesign. Computational benchmarks further demonstrated the framework's applicability to cyclic peptides and ligand-conditioned pocket design. TorchCraft extends predictor inversion to multiple binder formats and molecular contexts, providing a common framework for reusing all-atom structural priors in design.

cs.AI↗

arXiv2026

FactorEngine: A Program-level Knowledge-Infused Factor Mining Framework for Quantitative Investment

We study alpha factor mining, the automated discovery of predictive signals from noisy, non-stationary market data-under a practical requirement that mined factors be directly executable and auditable, and that the discovery process remain computationally tractable at scale. Existing symbolic approaches are limited by bounded expressiveness, while neural forecasters often trade interpretability for performance and remain vulnerable to regime shifts and overfitting. We introduce FactorEngine (FE), a program-level factor discovery framework that casts factors as Turing-complete code and improves both effectiveness and efficiency via three separations: (i) logic revision vs. parameter optimization, (ii) LLM-guided directional search vs. Bayesian hyperparameter search, and (iii) LLM usage vs. local computation. FE further incorporates a knowledge-infused bootstrapping module that transforms unstructured financial reports into executable factor programs through a closed-loop multi-agent extraction-verification-code-generation pipeline, and an experience knowledge base that supports trajectory-aware refinement (including learning from failures). Across extensive backtests on real-world OHLCV data, FE produces factors with substantially stronger predictive stability and portfolio impact-for example, higher IC/ICIR (and Rank IC/ICIR) and improved AR/Sharpe, than baseline methods, achieving state-of-the-art predictive and portfolio performance.

cs.AI↗

arXiv2026

Bi-Level Routing and Sparse Spatial Attention based Multi-View BEV 3D Object Detection for Autonomous Driving

Bird's Eye View (BEV)-based multi-view 3D object detection suffers from challenges of computational complexity, multi-scale feature extraction, and efficiency of dense 2D-to-BEV view transformation. To address these problems, this paper proposes an improved BEV 3D object detection algorithm Sparse-BEVNet. Firstly, a Bi-Level Routing Attention (BRA) mechanism is introduced into the image feature extraction network to reduce the computational burden of the backbone. Second, Cascaded Group Attention (CGA) is employed in the feature fusion module, which enhances deep interaction across features of different hierarchical levels without introducing additional computational overhead. Furthermore, a Sparse Spatial Cross-Attention mechanism is adopted to replace the conventional dense view projection pipeline. Experimental results on the public nuScenes dataset demonstrate that the proposed method achieves a mean Average Precision (mAP) of 45.2% and a nuScenes Detection Score (NDS) of 54.5%, corresponding to 3.6% and 2.8% improvements relative to the baseline model, respectively.

cs.CV↗

arXiv2026

Large-Time Asymptotics for the Kadomtsev-Petviashvili I Equation

We prove large time asymptotics for solutions of the KP I equation with small initial data. Our assumptions on the initial data rule out lump solutions but give a precise description of the radiation field at large times. Our analysis uses the inverse scattering method and involves large-time asymptotics for solutions to a non-local Riemann-Hilbert problem.

math.AP↗

arXiv2026

Sharp edge-spectral supersaturation for odd cycles

Let $G$ be a graph with $m$ edges and adjacency spectral radius$ρ(G)$, and let $N(C_{2k+1},G)$ denote the number of copies of $C_{2k+1}$ in $G$. For each fixed integer $k\ge 2$, define $ g_k(m):=\frac{k-1+\sqrt{4m-k^2+1}}{2}. $ Li, Zhai and Shu [European J. Combin., 2024] determined the spectral extremal threshold for odd cycles by proving that, for all sufficiently large $m$, every $C_{2k+1}$-free graph $G$ with $m$ edges satisfies $ρ(G)\le g_k(m)$. We establish the asymptotically sharp supersaturation counterpart of their result. More precisely, for every fixed integer $k\ge 2$, we prove that \[ \inf_{\substack{e(G)=m\\ ρ(G)>g_k(m)}} \frac{N(C_{2k+1},G)}{m^k} = \frac{\lceil k^2/2\rceil (k-1)!}{(k+1)^k}+o(1) \qquad\text{as }m\to\infty. \] Thus every $m$-edge graph whose spectral radius exceeds the $C_{2k+1}$-free threshold contains at least \[ \Big( \frac{\lceil k^2/2\rceil (k-1)!}{(k+1)^k}-o(1) \Big)m^k \] copies of $C_{2k+1}$, and the leading constant is asymptotically best possible. In particular, taking $k=2$, we obtain if $ ρ(G)>\frac{1+\sqrt{4m-3}}{2}$ then $ N(C_5,G)\ge \left(\frac{2}{9}-o(1)\right)m^2, $ with the constant $2/9$ being asymptotically optimal. This answers a question of Chen, Li and Tang concerning the existence and the largest possible value of a constant $C>0$ for which the same spectral condition guarantees at least $Cm^2$ copies of $C_5$. More generally, our result resolves a recent problem of Li, Lin, Liu and Zhang on spectral supersaturation for odd cycles. The proof combines spectral stability and resolvent analysis with estimates for odd spectral moments and a careful treatment of non-injective closed walks.

math.CO↗

arXiv2026

Separating Capability from Confidence: Grounded Dual-State Calibration for GRPO-Trained Medical Vision-Language Models

Medical vision-language models (VLMs) require confidence that reflects both answer correctness and patient-specific visual evidence. Recent GRPO-based methods optimize verbalized confidence together with answer generation. However, this joint optimization may interfere with answer learning and drive confidence toward near-binary values. Verbalized confidence also provides no explicit assessment of visual support. We therefore separate capability learning from confidence estimation and propose \textbf{DualRead}. DualRead builds on the insight that reliability can be read from the actor's internal states at critical moments in the answering process. It freezes the GRPO-trained actor and combines pre-answer solvability with a post-answer assessment of the generated answer and its visual support. To further assess whether confidence reflects visual grounding, we introduce \textbf{Counterfactual Confidence Grounding AUC} (CCG-AUC). It measures whether confidence decreases when real-image substitution changes the actor from correct to incorrect. Across two VLM backbones and both in- and out-of-distribution medical VQA benchmarks, DualRead improves correctness discrimination and calibration over verbalized confidence while preserving answer accuracy. CCG-AUC reveals whether confidence responds to answer-relevant visual evidence rather than primarily to non-visual cues.

cs.CV↗

arXiv2026

SVP Is NP-Hard for Some Rank-2 Cyclotomic Modules

Let $q$ range over primes congruent to $3$ modulo $4$. Let $ζ_q$ be a primitive $q$th root of unity, and put $K=\mathbb{Q}(ζ_q)$, with ring of integers $\mathcal{O}_K=\mathbb{Z}[ζ_q]$. We prove that the decision version of the Shortest Vector Problem ($\mathrm{SVP}$) in the $\ell_2$-norm is $\mathrm{NP}$-complete on full-rank free submodules of $\mathcal{O}_K^2$ by a deterministic polynomial-time many-one reduction from Exact Cover by 3-Sets (X3C). The module rank is fixed at two. As a $\mathbb{Z}$-lattice, the module has rank $2(q-1)$, which grows with $q$. The main obstacle is closure under the action of $\mathcal{O}_K$. A module containing a nonzero vector also contains every scalar multiple of that vector by a nonzero element of $\mathcal{O}_K$, and some of these multiples may be shorter. Three ideas overcome this obstacle. First, we map the Bennett--Peikert Reed--Solomon lattice to a principal cyclotomic ideal and use Wan's point-count estimates to prove that a coset of this ideal contains many binary coefficient representatives. Second, a checker based on a quadratic Gauss sum turns the X3C equations into a canonical squared norm. Third, the checker and a second module coordinate combine with a separation bound for ideal cosets to rule out every unintended vector created by the $\mathcal{O}_K$-action. Each constructed instance consists of a prime $q\equiv3\pmod4$, two integral generators whose $2\times2$ generator matrix has nonzero determinant, and an integer squared threshold. The construction also gives $\mathrm{NP}$-hardness of search-$\mathrm{SVP}$ under polynomial-time Turing reductions.

cs.CC↗

arXiv2026

Drift Calibration in Geometric Eye Tracking Systems

Geometric eye trackers can provide the spatial accuracy required for gaze-based interaction and multimodal studies, but their measurements remain sensitive to residual session-specific calibration error. Research on correcting this error is difficult to compare because methods are typically evaluated with different devices, target layouts, and error definitions. We present a calibration-focused dataset containing 163 trials from 12 participants, with separate 18-point fitting and 32-point test grids, and use it to evaluate global, local, and composite correction functions under a common spatial-extrapolation protocol. We further introduce a lightweight neural refiner that combines ranked predictions from complementary calibrators. On this controlled dataset, post-vendor correction reduces the mean angular error from $1.53^\circ$ to $1.03^\circ$ with the strongest classical composite and to $0.96^\circ$ with the refiner. In a closed-loop gaze task, lower residual error is associated with higher performance across four online correction conditions. These results provide a reproducible data-quality benchmark for using gaze as a behavioral signal in interactive modeling.

cs.CV↗

arXiv2026

Giant bulk photovoltaic effect driven by interfacial symmetry breaking in MoS2/Ta2NiSe5 heterostructures

Van der Waals (vdW) heterostructures offer a versatile platform for engineering unconventional bulk photovoltaic (BPV) effect through interfacial symmetry breaking. However, the coexistence of multiple photophysical mechanisms, driven by structural complexity, spontaneous charge transfer, and strong interlayer coupling, often obscures the microscopic origin of the BPV response and hinders its rational optimization. Here, we demonstrate a pronounced BPV effect localized at the overlap region of a cross-bar MoS2/Ta2NiSe5 vdW heterostructure, where symmetry breaking induced by vertical stacking lifts the inversion center of MoS2. The orthogonal device geometry enables the independent probing of intralayer and interfacial photoresponse pathways, facilitating clear separation of competing mechanisms. Spontaneous interfacial charge transfer between MoS2 and Ta2NiSe5 further establishes a strong interlayer electronic coupling. By modulating the interlayer potential landscape through gate voltage and vertical electric fields, we achieve an optimized zero-bias photocurrent density of 247 A/cm2 and a BPV coefficient of 0.99 V-1. Supported by theoretical modelling, our results illustrate how minimalist device geometry can transform complex heterostructures into experimentally tractable platforms. This strategy paves the way for analyzing and optimizing interface-driven BPV effect, with implications for self-powered optoelectronics, broadband photodetection, and energy-harvesting nanodevices.

cond-mat.mes-hall↗

arXiv2026

Intracranial Aneurysm Classification and Segmentation via Tri-Axial ROI and Multi-Task Learning

Intracranial aneurysms are often asymptomatic until rupture, which carries high mortality. Rupture risk assessment and treatment planning depend on both aneurysm morphology and anatomical location, yet existing automated methods remain limited to binary detection without fine-grained anatomical classification or multi-class segmentation. We present a multi-task framework that simultaneously performs multi-label classification, multi-class aneurysm segmentation, and multi-class vessel segmentation across 13 anatomical locations and four imaging modalities (CTA, MRA, T2, T1-post). Our two-stage approach combines a fast 2D tri-axial Region of Interest (ROI) extraction method with a 3D multi-task nnU-Net backbone. A dual-decoder design mitigates the extreme volume imbalance between aneurysm and vessel classes, while cross-attention pooling and modality-specific auxiliary heads improve feature learning across heterogeneous inputs. Our two-fold ensemble achieved 2nd place in the RSNA 2025 Intracranial Aneurysm Detection challenge. Code, model weights, a 3D Slicer plugin, and the corrected segmentation labels are publicly available.

cs.CV↗

arXiv2026

SkyNative: A Native Multimodal Architecture for Remote Sensing Vision-Language Understanding

Remote sensing vision-language models (RS-VLMs) commonly employ a pretrained vision encoder and a projection module to map image features into the token space of a large language model. Although effective, this modular RS-VLMs separates visual representation from language reasoning, potentially limiting the direct involvement of fine-grained visual evidence in complex spatial inference. This challenge is particularly relevant to remote sensing imagery, which often covers broad geographic areas and contains multi-scale objects, dense target distributions, and intricate spatial layouts. In this paper, we propose SkyNative, the first study to explore a native multimodal architecture for remote sensing vision-language tasks. SkyNative converts remote sensing images into visual tokens through a lightweight patch embedding module and places them together with text tokens in a shared autoregressive sequence, allowing textual tokens to directly access the preceding visual context. To accommodate the heterogeneous characteristics of the two modalities, we further adopt a modality-aware decoupling mechanism that applies modality-specific projections, normalization, and feed-forward transformations while processing both modalities through shared causal self-attention. Extensive experiments demonstrate SkyNative's strong capabilities in dense small-object perception, large-format contextual understanding, complex reasoning, and robustness, with scores of 68.93%, 47.40%, and 63.43% on HRRSD, RSHR reasoning, and OmniEarth, respectively. These results suggest that the native VLM architecture explored in SkyNative represents a promising approach to RS vision-language modeling.

cs.CV↗