arXiv subjects

Lei Li

Publications and source records attributed to Lei Li.

At least 19 recordsLinked to original sources

arXiv2026

On the propagation of chaos for three-dimensional Coulomb Brownian particles

While the mean field limit of the Brownian particles with 3D Coulomb interaction has been resolved starting from the Liouville equation, the justification directly from the random trajectories is not fully resolved. In our previous work (J. Math. Phys., 60, 111501, 2019), it has been shown that the time marginals of any limit point of the empirical measure form a weak solution to the mean field nonlinear Fokker-Planck equation. However, the identification to the strong solution is not established due to the lack of regularity. In this short note, we first make use of the variational formulation of the quadratic function to establish the desired space-time $L^2$ bound mentioned already in J. Math. Phys., 60, 111501, 2019. This fills gap and shows that the time marginals must be the must be the strong solution. Moreover, making use of the virial type estimates and the force dissipation, we also establish the propagation of chaos in path space directly from the random trajectories of the particle systems.

math.PR↗

arXiv2026

DiffCap-Bench: A Comprehensive, Challenging, Robust Benchmark for Image Difference Captioning

Image Difference Captioning (IDC) generates natural language descriptions that precisely identify differences between two images, serving as a key benchmark for fine-grained change perception, cross-modal reasoning, and image editing data construction. However, existing benchmarks lack diversity and compositional complexity, and standard lexical-overlap metrics (e.g., BLEU, METEOR) fail to capture semantic consistency or penalize hallucinations, which together prevent a comprehensive and robust evaluation of multimodal large language models (MLLMs) on IDC. To address these gaps, we introduce DiffCap-Bench, a comprehensive IDC benchmark covering ten distinct difference categories to ensure diversity and compositional complexity. Furthermore, we propose an LLM-as-a-Judge evaluation protocol grounded in human-validated Difference Lists, enabling a robust assessment of models' ability to both capture and describe visual changes. Through extensive evaluation of state-of-the-art MLLMs, we reveal significant performance gaps between proprietary and open-source models, highlight the critical importance of reasoning capability, and identify clear limitations in model scaling. Our framework also demonstrates strong alignment with human expert judgments and strong correlation with downstream image editing data construction quality. These findings establish DiffCap-Bench as both a reliable IDC evaluation framework and a practical predictor of downstream utility. The benchmark and code will be made publicly available to support further research.

cs.CV↗

arXiv2026

Minimax regret and monotone optimality in multiple testing

Confirmatory trials may seek efficacy claims in a prespecified subgroup and the full popu- lation, although benefit outside the subgroup is uncertain. We measure power loss relative to the marginal level-α test available if the population to protect were known, using the same experiment. In a bivariate Gaussian model with known covariance, we minimise the largest power loss, or regret, over prespecified ranges of target and non-target effects, subject to strong familywise error control and coordinatewise monotonicity. We optimise over monotone closed- testing procedures with fixed marginal tests, allowing randomisation in the shared intersection test. Monotonicity places the worst non-target mean at its lower planning bound for each target effect. On the resulting reduced planning set, every least-favourable distribution is supported on exactly two efficacy scenarios, one for each target population. Under a condition on the effect ranges and covariance, the minimax procedure is nonrandomised and unique up to null sets. An integral criterion characterises exactly when a calibrated step-up procedure attains the minimax value within this class. Bounds for the continuous problem quantify possible improve- ments over simple procedures. An asthma planning example illustrates power allocation and the compatibility of target assignments with treatment benefits and burdens.

stat.ME↗

arXiv2026

VoxLution: Evolving Voxels for High-Fidelity 3D Scene Generation

High-resolution 3D indoor scene generation is crucial for immersive applications such as virtual reality, spatial computing, and robotics simulation. We introduce VoxLution, a coarse-to-fine transformer-based approach that generates high-fidelity indoor scenes with rich details, overcoming the overly smooth geometry typical of existing generative models. First, we generate a low-resolution spatial layout; then, based on this structural information, we progressively refine the scene geometry locally in chunks. Key to our approach is formulating 3D super-resolution as a sequence modeling problem over multi-resolution sparse voxel grids. To achieve this, we introduce a training-free, resolution-invariant tokenization scheme that compactly encodes the geometric information of eight octants into a single decimal number. We formulate the transformer sequence to include multiple resolutions, allowing the coarse layout to guide the generation of fine local details. To ensure coherence across local chunks, we incorporate adjacent chunk boundary information directly into the transformer sequencing. VoxLution achieves scene generation at a finer voxel resolution than existing baselines, demonstrating state-of-the-art performance on standard benchmarks with 59.6% lower FPD and 51.1% lower FID on 3D-FRONT, and 70.5% lower FPD and 73.8% lower FID on Scannet++.

cs.CV↗

arXiv2026

When Requirements Interact: Scaling Long-Horizon Coding via Interdependent Specifications

To succeed in real-world long-horizon software development, coding agents must satisfy multi-dimensional task specifications, such as interface compatibility and error handling, and do so without breaking existing dependencies. However, current works lack specification in both quantity and diversity, often emphasizing primary goals while ignoring complex constraints. Synthesizing tasks to bridge this gap presents two challenges: naively adding specifications risks reducing tasks to trivial checklists rather than compelling joint reasoning, and verifying these subtle interactions risks overlooking hidden flaws or rejecting valid alternatives. To address these challenges, we introduce CodeSpecWeave, a framework for synthesizing tasks with rich, interdependent specifications. Grounded in real developer seeds, a synthesis agent iteratively develops both coupled task requirements and executable tests, while a review agent audits solution trajectories from multiple models to guarantee test reliability. Using this pipeline, we construct 4,670 training tasks across 2,865 repositories. Applying Reinforcement learning on these tasks significantly improves performance over the base model on DeepSWE-v1.1, largely outperforming existing baselines. Furthermore, we curate CodeSpecWeave-Bench, a rigorous benchmark comprising 100 tasks and 1,044 specifications, which clearly separates frontier models and reveals significant headroom in handling interdependent engineering constraints.

cs.SE↗

arXiv2026

A discrete fractional Grönwall inequality for dissipative systems via resolvent kernel estimate

We establish a discrete fractional Gronwall inequality for dissipative systems with time-dependent forcing. For completely positive discretizations, we bound individual resolvent entries by integrals of the continuous response kernel, under controlled perturbation to the L1 weights and a stepsize restriction. The proof constructs a supersolution for each resolvent column. The resulting Mittag-Leffler weights retain both the decay of initial data and the historical contributions of forcing. The framework covers L1 and several standard schemes on uniform meshes. For linear subdiffusion, it yields decay-preserving L1 error estimates separating first order and $(2-α)$ order contributions, with stronger long-time decay under additional solution regularity.

math.NA↗

arXiv2026

Whether or Where? Tail Calibration Signals Finite-Budget Reward Hacking in Best-of-N Selection

Best-of-N (BoN) inference returns the candidate with the highest proxy reward, so a larger $N$ tests the proxy's extreme tail more than its average ranking quality. We show that the exact BoN value is a Beta-weighted integral of conditional true reward over within-prompt score quantiles. It follows that BoN depends on scores only through their ranks. It also follows that BoN can decline only if the calibration curve in the tail turns downward, and the budget where the decline appears depends on where that turn lies. Under a bounded additive model, the error tail sets the limit: Gaussian errors keep improving, Laplace errors reach a plateau, and heavy-tailed errors return to the prior mean. When error variance is largest at low true reward, even Gaussian noise can push BoN below random selection. Synthetic phase maps agree with this taxonomy. On real candidate pools spanning several verifiable domains, base policies, and learned reward models, tail failures are rare and all occur in code generation. A held-out tail calibration statistic reliably predicts where a cross-fitted budget cap helps. It does about as well as rank accuracy, and the cap is competitive with a Soft-BoN baseline. Tail calibration therefore does not replace ranking metrics. It is an audit that is unchanged by any monotone rescaling of the reward scores and shows which budgets are at risk.

cs.LG↗

arXiv2026

Ulysses: A Security Harness For Synthesizing Verifiable Web Exploits

How can an AI agent automatically discover security vulnerabilities in any software repository? Existing approaches use LLMs to generate software vulnerability exploits. However, it is challenging for an LLM to reliably verify its own generated exploits. In this paper, we propose Ulysses, a multi-agent security harness to synthesize verifiable vulnerability exploits for web applications. Ulysses converts candidate vulnerability findings into verified Proofs of Exploitability by iteratively exploring exploit hypotheses and establishing success in independently observed execution effects. We evaluated Ulysses on WebExploitGym, a curated benchmark of exploitable instances from 25 real-world web applications that span 5 languages and 7 frameworks. Ulysses produces verified proofs for 10 of 25 real-world advisories. It also discovered 4 new proofs of exploitability that have never been previously reported.

cs.CR↗

arXiv2026

Self-Evolving Static Analysis via LLM Pseudo-Labeling

Security bugs often slip past static analysis not because they are impossible to detect, but because detection rules have not yet been written for the code in front of us. New frameworks, APIs, and coding idioms continually create such gaps, leaving developers to discover failures and manually encode new rules. We present SESAA, a framework in which an Large Language Model (LLM) agent starts from no rules and iteratively evolves a set of readable and auditable joern queries. The agent receives feedback only from a second LLM critic, which inspects flagged and unflagged code to identify likely false alarms and suspected misses, without access to ground-truth labels. The agent then uses this feedback to revise its queries. On the OWASP Benchmark for Python, SESAA reaches a mean F1 of 0.742 across 8 independent runs, well above the default configurations of five widely used tools and comparable to the strongest of them once it is tuned with access to ground-truth labels that SESAA never sees. In a pilot on real-world Python projects, queries induced from code available before a vulnerability was introduced detect 10 of 19 subsequent CVEs, including 9 detected by rules written for the same repository. The resulting detector is a set of executable, human-readable queries that engineers can inspect, modify, and run independently. These results suggest that LLM agents can turn static-analysis rule maintenance from a manually authored process into an iterative process of automated discovery and repair.

cs.CR↗

arXiv2026

Quasi-random splitting method for accurate and efficient multiphysics simulation

We propose a quasi-random operator splitting method for evolution equations driven by multiple mechanisms. The method uses a one-dimensional low-discrepancy sequence to determine the ordering of the subflows through a direct point-to-permutation construction, while applying each subflow exactly once per time step. Thus, for a decomposition into $p$ operators, the proposed method requires only $p$ subflow evaluations per step, compared with essentially $2p-2$ evaluations for the classical multi-operator Strang splitting when $p\geq 3$. In contrast to randomized splitting, the quasi-random ordering is deterministic once the underlying sequence is fixed and avoids the run-to-run fluctuations caused by random permutations. To analyze the resulting scheme, we develop a convergence framework that connects the discrepancy of the induced operator orderings with cancellation of the leading commutator errors accumulated over time. For an arbitrary fixed number of bounded linear operators, we establish a global error of order $O(τ^2\log N)$, equivalently $O(τ^2|\logτ|)$ for fixed terminal time, which is essentially second order. We further extend the same cancellation mechanism to the nonlinear Allen--Cahn equation and obtain the same essentially second-order convergence behavior under suitable Sobolev regularity assumptions. Numerical experiments for bounded linear systems and the Allen--Cahn equation confirm the predicted convergence rates and show that the proposed quasi-random splitting strategy retains the low per-step computational cost of a Lie-type splitting method while achieving accuracy close to second order.

math.NA↗

arXiv2026

LCPNet: Latent Consistent Proximal Unfolding Network for Infrared Small Target Detection

Infrared small target detection (IRSTD) aims to identify long distance small targets from complex infrared backgrounds, and is a fundamental task in remote sensing. Deep learning methods have improved IRSTD by learning discriminative image-to-mask mappings, but such feed-forward designs often underuse physical decomposition structure between targets and backgrounds. Deep unfolding methods partially address this issue by embedding model-driven iterations into neural networks, yet existing designs still operate mainly in image domain and use updates and memory mechanisms that are not fully coupled with underlying optimization process. To address these limitations, we propose Latent Consistent Proximal unfolding network (LCPNet). First, we verify that low-rank prior remains valid in latent representations and perform unfolding in this space, preserving physical constraint while avoiding repeated compression of intermediate states. Second, we derive a Latent Consistent Proximal (LCP) solver that evolves each latent variable from its previous state rather than reconstructing through an indirect residual, and stabilizes small target updates through task-adaptive normalization and gain control. Third, we introduce Shared Optimization Memory (SOM), a common historical state shared by all decomposition variables to provide coordinated guidance across unfolding stages. Extensive experiments on four public benchmarks demonstrate that LCPNet achieves accurate and robust detection with low false-alarm rates, together with competitive efficiency among high-accuracy deep unfolding methods. Model and code are available at Model and code are available at https://github.com/Tianfang-Zhang/LCPNet.

cs.CV↗

arXiv2026

InstructPro2: Natural Language and Biological Grounding Guided De Novo Protein Design

The ability to generate proteins with desired functions is a central goal of computational biology. Although natural language can express diverse design objectives, existing protein generation methods often support only narrow tasks or make limited use of biological constraints. We develop InstructPro2 to translate natural-language design objectives into protein sequences while incorporating task-specific biological context. A text encoder captures the design instructions, while dedicated ligand and antigen encoders provide biological context to the sequence-generating decoder. For pretraining, we assemble InstructPro2Bench, comprising 205.7 million pairs from UniProtKB. We subsequently adapt the model to ligand-binding protein and antibody design. On functional protein generation benchmarks, InstructPro2 outperforms Pinal, the strongest baseline, achieving a functional recovery F1 of 74.8 versus 62.2 and lowering global RMSD from 5.33A to 3.04A. For antibody design, the model obtains a DockQ score of 0.39, while ligand-binding protein design reaches a docking success rate of 72.6%. These findings highlight how integrating language-based instructions with biological conditioning can extend protein generation to distinct design settings.

cs.LG↗

arXiv2026

Joint Discrete and SE(3) Diffusion for Protein Interface Design

The design of high-affinity, target-specific protein binders is central to the development of modern therapeutics. However, existing protein-generation methods remain limited in this setting, as many generated designs fail to simultaneously achieve geometric and physicochemical complementarity at binding interfaces. To address this challenge, we introduce PIDiff, a physicochemically enhanced multimodal diffusion framework for protein-protein complex design. PIDiff jointly designs target-conditioned protein sequences and structures by coupling diffusion in discrete sequence space with diffusion on the continuous (SE(3)) structural manifold. At its core, PIDiff employs PIFormer, a novel architecture that interleaves geometry-aware invariant point attention with global sequence self-attention. We further curate PIBench, a protein-protein complex dataset with diverse target scaffolds derived from the Protein Data Bank (PDB). We first pretrain PIDiff on PIBench and subsequently fine-tune it for two therapeutically relevant applications: minibinder design and antibody design. PPFrame consistently outperforms baseline methods across sequence design success rate, structure binding energy, and sequence-structure co-designability on all tasks.

cs.AI↗

arXiv2026

CoGameReg: Cooperative-Game-Based Collaborative Semi-Supervised Regression for Seismic Reservoir Property Prediction

Accurate estimation of reservoir properties from seismic data remains challenging when well-log labels are sparse but three-dimensional seismic observations are abundant. We present CoGameReg, a cooperative-game-based semi-supervised regression framework that uses contribution-weighted predictions of heterogeneous regressors to construct shared pseudo-labels. This design makes each regressor's influence explicit in the learning targets generated for unlabeled seismic samples under sparse well control. The Tri-Game Shapley mechanism combines a rank-MSE performance game with a pairwise consistency game to estimate nonnegative regressor weights from labeled training data. These weights remain fixed during collaborative training and are reused for final prediction fusion. Candidate pseudo-labels are screened using a common local k-nearest-neighbor error-reduction criterion before retraining. We evaluate CoGameReg on 10 public regression datasets, a synthetic three-dimensional P-wave velocity prediction task, and a field water-saturation application. The benchmark advantage is most pronounced at the 2.5% labeled-data ratio and decreases as supervision increases. CoGameReg obtains the best blind-well metrics in the synthetic experiment. In the field application, predictions clipped to 0-100% yield a pooled coefficient of determination of 0.5942, a root mean square error of 5.7441 percentage points, and a Pearson correlation of 0.8265 across three blind wells. CoGameReg has the highest Pearson correlation among the four field methods. The results support the use of CoGameReg to integrate sparse well labels with dense seismic attributes, with performance varying across metrics and well locations.

cs.LG↗

arXiv2026

MiMo-V2.6: Scaling Reinforcement Learning Towards Self-Improvement

Reinforcement learning (RL) is the central training paradigm for advancing large foundation models towards self-improvement. This report introduces the MiMo-V2.6 series, an omni-modal family that pushes the frontier of model intelligence by scaling RL compute. Prior to RL, we conduct mid-training on a broad multimodal corpus to provide ample exploration space, and build a solid infrastructure on the pretrained hybrid-SWA architecture to support subsequent scale-up. We scale RL compute along three dimensions: (1) larger batches and higher throughput, with an asynchronous training that consumes 1,568 samples and 2.7-3.7B tokens per step at context lengths of up to 1M; (2) more diverse and complex environments, spanning code, general, visual, and cyber domains under a mixture of agent harnesses; and (3) more grader compute, via groupwise agentic grading that yields more accurate reward signals for long-horizon tasks and steers the model towards shorter, more token-efficient solutions. To keep training stable at scale, we freeze the MoE router and establish a multi-layer defense against reward hacking. We further build infrastructure for mixed-task agentic RL, including a unified trajectory representation, high-concurrency multi-framework rollout, decoupled control and data planes, and training-inference consistency. We open-source the training dynamics, RL environments, and RL framework to facilitate reproduction and further research on scaled RL and model self-improvement.

cs.CL↗

arXiv2026

Is the Future Predictable? Modeling Future Predictability for Reliable Temporal Reasoning

Large language models (LLMs) remain unreliable when reasoning about the future. Existing models often overlook the difference among future states, committing to specific answers even when the queried facts are inherently volatile. To capture this challenge, we introduce ChronoRegime, a future-oriented temporal reasoning task that requires models to determine whether historical temporal evidence justifies commitment to a future answer. Furthermore, we propose Confidence-Aware Temporal Reasoning (CTR), a framework that estimates future predictability from temporal dynamics and regulates answer commitment accordingly. Rather than performing unconditional forecasting, CTR commits only when the reconstructed evidence supports the target future state. Across multiple LLMs, CTR improves strict overall accuracy by 6.7 percentage points and reduces hallucination rates by 18.4 percentage points on average over the plain QA baseline. These results demonstrate that explicit predictability assessment is an effective mechanism for safer and more reliable temporal reasoning under future uncertainty.

cs.CL↗

arXiv2026

AuraForge: Scaling Security Supervision for Training Coding Agents

Coding agents are now proficient enough to generate complex software applications from a single prompt. As their capabilities have grown, human oversight has increasingly shifted from line-by-line code review toward hands-off evaluation of outcomes. However, recent studies have shown that such a transition exposes a critical risk: functional correctness alone does not guarantee a secure implementation. Despite growing attention to code security, training safer coding agents remains challenging because reliable security supervision is difficult to obtain at scale from real-world repositories. We introduce AuraForge to synthesize and validate executable security tests for training secure coding agents. Our approach combines attack-oriented test synthesis, language-extensible task construction, and safeguards against reward hacking. Using AuraForge, we construct AuraGym, a multi-language and multi-CWE executable training gym: 679 executable feature-implementation tasks from 344 real-world repositories across Python, JavaScript, and TypeScript, covering 177 CWE categories. On the subset with human-written security tests, AuraForge produces about 3 times as many test cases on average and reduces the false-positive rate by 83.23%, allowing alternative secure implementations to receive correct supervision. Training Qwen3.5-4B with synthesized security tests gains larger improvements than human-written security tests (average 19.7 FuncPass and 6.2 SecPass vs. 14.9 FuncPass and 4.4 SecPass) on three languages. These results demonstrate that AuraForge provides more diverse and reliable security supervision to train secure coding agents.

cs.CR↗

arXiv2026

ABSENTIA: Detecting Broken Access Control Vulnerabilities in Web Applications

Broken access control, the failure of authorization, is one of the most prevalent web security risks. Unlike injection, a flow of untrusted input into a dangerous operation, authorization is a relation: who may act on what, not how data moves. Each application decides that relation for itself, so no rule written in advance carries to the next. An LLM agent can infer it from the code, but with no systematic way to cover the application and prioritize what to inspect, its search stays undirected and access-control flaws go undetected. We present ABSENTIA, a security scaffolding that turns general LLM agents into systematic vulnerability detectors for the backend of web applications, run as an audit by the developers and security engineers who maintain the code. Under its direction, the agents build a graph that maps the application's routes to the code behind them. ABSENTIA then works route by route, applying invariant falsification: it infers the properties the code is meant to satisfy, and where one is not enforced, reports the route for maintainer review. We also release BAC-Bench, a benchmark of 30 broken access control advisories across 25 repositories, 3 languages, and 9 frameworks, each published in 2025 or later, verified by a human auditor, and paired with its fixing commit, so credit requires flagging the vulnerable version and not the fixed one. ABSENTIA recalls 19 of them, 17 under paired credit, and an LLM verifier confirms 51% of its findings. CodeQL and Semgrep recall none, and an unstructured agent on the same model recalls 3. In the OWASP Benchmark injection categories, ABSENTIA leads the dedicated analyzers in Python and trails only CodeQL and IRIS in Java.

cs.CR↗