arXiv subjects

Xiaolong Li

Publications and source records attributed to Xiaolong Li.

At least 37 records · Page 2Linked to original sources

arXiv2026

The Script is All You Need: An Agentic Framework for Long-Horizon Dialogue-to-Cinematic Video Generation

Recent advances in video generation have produced models capable of synthesizing stunning visual content from simple text prompts. However, these models struggle to generate long-form, coherent narratives from high-level concepts like dialogue, revealing a ``semantic gap'' between a creative idea and its cinematic execution. To bridge this gap, we introduce a novel, end-to-end agentic framework for dialogue-to-cinematic-video generation. Central to our framework is ScripterAgent, a model trained to translate coarse dialogue into a fine-grained, executable cinematic script. To enable this, we construct ScriptBench, a new large-scale benchmark with rich multimodal context, annotated via an expert-guided pipeline. The generated script then guides DirectorAgent, which orchestrates state-of-the-art video models using a cross-scene continuous generation strategy to ensure long-horizon coherence. Our comprehensive evaluation, featuring an AI-powered CriticAgent and a new Visual-Script Alignment (VSA) metric, shows our framework significantly improves script faithfulness and temporal fidelity across all tested video models. Furthermore, our analysis uncovers a crucial trade-off in current SOTA models between visual spectacle and strict script adherence, providing valuable insights for the future of automated filmmaking.

cs.CV↗

arXiv2026

Adaptive Preconditioners Trigger Loss Spikes in Adam

Loss spikes commonly emerge during neural network training with the Adam optimizer across diverse architectures and scales, yet their underlying mechanism remains elusive. While previous explanations attribute these phenomena to sharper loss landscapes at lower loss, we show that landscape geometry alone is insufficient to explain the phenomenon. In this work, we pinpoint the root cause in the internal dynamics of Adam's second moment estimator. We identify a critical ``decoupling'' mechanism where the adaptive preconditioner $v_t$ fails to track the instantaneous squared gradients $g_t^2$, causing the adaptive mechanism to effectively fail. This decoupling allows the preconditioner to decay autonomously despite rising gradients, which pushes the maximum eigenvalue of the preconditioned Hessian beyond the stability threshold $2/η$ for sustained periods, manifesting as dramatic loss spikes. Through a quadratic approximation analysis, we theoretically and experimentally characterize five distinct stages of spike evolution and propose a predictor for anticipating spikes based on gradient-directional curvature. We empirically find that the proposed loss spike mechanism, although derived from simplified models, generalizes well to practical scenarios ranging from small neural networks to large-scale Transformers.

cs.LG↗

arXiv2026

Discovering Physical Directions in Weight Space: Composing Neural PDE Experts

Recent advances in neural operators have made partial differential equation (PDE) surrogate modeling increasingly scalable and transferable through large-scale pretraining and in-context adaptation. However, after a shared operator is fine-tuned to multiple regimes within a continuous physical family, it remains unclear whether the resulting weight-space updates merely form isolated regime experts or reveal reusable physical structure. Starting from a shared family anchor, we fine-tune low- and high-regime endpoint experts and show that their updates can be separated into a family-shared adaptation and a direction aligned with the underlying physical parameter. This separation reinterprets endpoint experts as finite-difference probes of a local physical direction in weight space, explaining why static averaging can interpolate between regimes but attenuates endpoint-specific physics. Building on this perspective, we propose Calibration-Conditioned Merge (CCM), a post-hoc coordinate readout method for composing neural PDE experts along this physical direction. Given physical metadata, a calibrated coordinate mapping, or a short observed rollout prefix, CCM infers the target composition coordinate and deploys a single merged checkpoint for the remaining rollout. We evaluate CCM on the reaction--diffusion system, viscosity-parameterized two-dimensional Navier--Stokes equations, and radial dam-break dynamics. Across these benchmarks, CCM achieves its strongest gains in extrapolative regimes, reducing out-of-distribution rollout error relative to the family anchor by 54.2%, 42.8%, and 13.8%, respectively. Further experiments across FNO scales, a DPOT-style backbone, and ablations confirm that endpoint fine-tuning is not arbitrary checkpoint drift, but reveals a calibratable physical direction for training-free transfer across PDE regimes.

cs.LG↗

arXiv2026

Expanding the High-z Supernova Frontier: "Wide-Area" JWST Discoveries from the First Two Years of COSMOS-Web

Transient astronomy in the early Universe (z > 2) remains largely unexplored, lying beyond the rest-frame optical spectroscopic reach of most current observatories. Yet this regime promises transformative insights, with high-redshift transients providing direct access to the early Universe and enabling studies of how stellar populations and cosmology evolve over cosmic time. JWST is uniquely equipped to probe these redshifts efficiently in the rest-frame optical and near-IR. We present results from an initial pathfinder search, covering an area of ~133 arcmin^2 (~0.037 deg^2) independently imaged by the PRIMER and COSMOS-Web (hereafter COSMOS) extragalactic surveys. Although neither program was designed for time-domain astronomy, combining their data results in difference images separated by roughly one year, leading to the discovery of 68 supernovae (SNe) with host photometric redshifts reaching z < 5. For most SNe, only a single epoch is available, but the combination of host redshift, classification, color, and magnitude enables us to prioritize candidates for detailed photometric and spectroscopic follow-up. Among the most notable sources are a relatively bright, blue CCSN at z > 3 (SN 2023aeab) and a young, normal SN Ia at z > 2 (SN 2023aeax). The sample distribution highlights the increasing likelihood that a wide-area JWST program can uncover younger, bluer, and potentially more extreme explosions. While this pathfinder effort is limited in cadence and number of filters, it demonstrates the strong potential of a dedicated, well-planned time-domain survey with JWST to obtain the sample sizes and rate measurements needed to chart SN populations deep into the early Universe.

astro-ph.HE↗

arXiv2026

Context-Fidelity Boosting: Enhancing Faithful Generation through Watermark-Inspired Decoding

Large language models (LLMs) often produce content that contradicts or overlooks information provided in the input context, a phenomenon known as faithfulness hallucination. In this paper, we propose Context-Fidelity Boosting (CFB), a lightweight and general decoding-time framework that reduces such hallucinations by increasing the generation probability of source-supported tokens. Motivated by logit-shaping principles from watermarking techniques, CFB applies additive token-level logit adjustments based on a token's degree of support from the input context. Specifically, we develop three boosting strategies: static boosting, which applies a fixed bias to source-supported tokens; context-aware boosting, which scales this bias using the divergence between next-token distributions with and without context; and token-aware boosting, which further redistributes the adaptive bias according to local relevance estimated from source-position attention and source-scoped semantic similarity. CFB requires no retraining or architectural changes, making it compatible with a wide range of LLMs. Experiments on summarization and question answering tasks across multiple open-source LLMs show that CFB consistently improves faithfulness metrics with minimal generation overhead. Our implementation is fully open-sourced.

cs.CL↗

arXiv2026

Non-existence of Lyapunov exponents in the Newhouse domain

We show that within the Newhouse domain of $C^r$ surface diffeomorphisms ($r \in [2,\infty )$), there exists a dense subset $\mathcal D$ such that for any $f \in \mathcal D$, Lyapunov exponents fail to exist for all points in some open set $U$ and all nonzero tangent vectors in some open cone $V \subset \mathbb{R}^2$. This demonstrates that the non-existence of Lyapunov exponents is a persistent phenomenon in the setting of robust homoclinic tangencies. The proof relies on constructing diffeomorphisms exhibiting specific oscillatory return times near a homoclinic tangency, incorporating techniques from Newhouse theory and recent results on Lyapunov irregularity, alongside several refinements and new arguments.

math.DS↗

arXiv2026

UCAgent: An End-to-End Agent for Block-Level Functional Verification

Functional verification remains a critical bottleneck in modern IC development cycles, accounting for approximately 70% of total development time in many projects. However, traditional methods, including constrained-random and formal verification, struggle to keep pace with the growing complexity of modern semiconductor designs. While recent advances in Large Language Models (LLMs) have shown promise in code generation and task automation, significant challenges hinder the realization of end-to-end functional verification automation. These challenges include (i) limited accuracy in generating Verilog/SystemVerilog verification code, (ii) the fragility of LLMs when executing complex, multi-step verification workflows, and (iii) the difficulty of maintaining verification consistency across specifications, coverage models, and test cases throughout the workflow. To address these challenges, we propose UCAgent, an end-to-end agent that automates hardware block-level functional verification based on three core mechanisms. First, we establish a pure Python verification environment using Picker and Toffee to avoid relying on LLM-generated SystemVerilog verification code. Second, we introduce a configurable 31-stage fine-grained verification workflow to guide the LLM, where each stage is verified by an automated checker. Furthermore, we propose a Verification Consistency Labeling Mechanism (VCLM) that assigns hierarchical labels to LLM-generated artifacts, improving the reliability and traceability of verification. Experimental results show that UCAgent can complete end-to-end automated verification on multiple modules, including the UART, FPU, and integer divider modules, achieving up to 98.5% code coverage and up to 100% functional coverage. UCAgent also discovers previously unidentified design defects in realistic designs, demonstrating its practical potential.

cs.SE↗

arXiv2026

BIRD-INTERACT: Re-imagining Text-to-SQL Evaluation for Large Language Models via Lens of Dynamic Interactions

Large language models (LLMs) have demonstrated remarkable performance on single-turn text-to-SQL tasks, but real-world database applications predominantly require multi-turn interactions to handle ambiguous queries, execution errors, and evolving user requirements. Existing multi-turn benchmarks fall short by treating conversation histories as static context or limiting evaluation to read-only operations, failing to reflect production-grade database assistant challenges. We introduce BIRD-INTERACT, a benchmark that restores this realism through: (1) a comprehensive interaction environment coupling each database with a hierarchical knowledge base, metadata files, and a function-driven user simulator, enabling models to solicit clarifications, retrieve knowledge, and recover from errors without human supervision; (2) two evaluation settings consisting of a pre-defined conversational protocol (c-Interact) and an open-ended agentic setting (a-Interact) where models autonomously decide when to query the user simulator or explore the environment; (3) a challenging task suite covering the full CRUD spectrum for business-intelligence and operational use cases, guarded by executable test cases. Each task features ambiguous and follow-up sub-tasks requiring dynamic interaction. The suite comprises BIRD-INTERACT-FULL (600 tasks, up to 11,796 interactions) for comprehensive performance assessment, and BIRD-INTERACT-LITE (300 tasks with simplified databases) for detailed behavioral analysis and rapid method development. Our empirical results highlight BIRD-INTERACT's difficulty: GPT-5 completes only 8.67% of tasks in c-Interact and 17.00% in a-Interact. Analysis via memory grafting and Interaction Test-time Scaling validates the importance of effective interaction for complex, dynamic text-to-SQL tasks.

cs.AI↗

arXiv2026

3D-Layout-R1: Structured Reasoning for Language-Instructed Spatial Editing

Large Language Models (LLMs) and Vision Language Models (VLMs) have shown impressive reasoning abilities, yet they struggle with spatial understanding and layout consistency when performing fine-grained visual editing. We introduce a Structured Reasoning framework that performs text-conditioned spatial layout editing via scene-graph reasoning. Given an input scene graph and a natural-language instruction, the model reasons over the graph to generate an updated scene graph that satisfies the text condition while maintaining spatial coherence. By explicitly guiding the reasoning process through structured relational representations, our approach improves both interpretability and control over spatial relationships. We evaluate our method on a new text-guided layout editing benchmark encompassing sorting, spatial alignment, and room-editing tasks. Our training paradigm yields an average 15% improvement in IoU and 25% reduction in center-distance error compared to Chain of Thought Fine-tuning (CoT-SFT) and vanilla GRPO baselines. Compared to SOTA zero-shot LLMs, our best models achieve up to 20% higher mIoU, demonstrating markedly improved spatial precision.

cs.CV↗

arXiv2026

BrainSCL: Subtype-Guided Contrastive Learning for Brain Disorder Diagnosis

Mental disorder populations exhibit pronounced heterogeneity -- that is, the significant differences between samples -- poses a significant challenge to the definition of positive pairs in contrastive learning. To address this, we propose a subtype-guided contrastive learning framework that models patient heterogeneity as latent subtypes and incorporates them as structural priors to guide discriminative representation learning. Specifically, we construct multi-view representations by combining patients' clinical text with graph structure adaptively learned from BOLD signals, to uncover latent subtypes via unsupervised spectral clustering. A dual-level attention mechanism is proposed to construct prototypes for capturing stable subtype-specific connectivity patterns. We further propose a subtype-guided contrastive learning strategy that pulls samples toward their subtype prototype graph, reinforcing intra-subtype consistency for providing effective supervisory signals to improve model performance. We evaluate our method on Major Depressive Disorder (MDD), Bipolar Disorder (BD), and Autism Spectrum Disorders (ASD). Experimental results confirm the effectiveness of subtype prototype graphs in guiding contrastive learning and demonstrate that the proposed approach outperforms state-of-the-art approaches. Our code is available at https://anonymous.4open.science/r/BrainSCL-06D7.

cs.LG↗

arXiv2026

TokenCom: Vision-Language Model for Multimodal and Multitask Token Communications

Visual-Language Models (VLMs), with their strong capabilities in image and text understanding, offer a solid foundation for intelligent communications. However, their effectiveness is constrained by limited token granularity, overlong visual token sequences, and inadequate cross-modal alignment. To overcome these challenges, we propose TaiChi, a novel VLM framework designed for token communications. TaiChi adopts a dual-visual tokenizer architecture that processes both high- and low-resolution images to collaboratively capture pixel-level details and global conceptual features. A Bilateral Attention Network (BAN) is introduced to intelligently fuse multi-scale visual tokens, thereby enhancing visual understanding and producing compact visual tokens. In addition, a Kolmogorov Arnold Network (KAN)-based modality projector with learnable activation functions is employed to achieve precise nonlinear alignment from visual features to the text semantic space, thus minimizing information loss. Finally, TaiChi is integrated into a multimodal and multitask token communication system equipped with a joint VLM-channel coding scheme. Experimental results validate the superior performance of TaiChi, as well as the feasibility and effectiveness of the TaiChi-driven token communication system.

cs.CV↗

arXiv2026

Mechanically Assisted Symmetry Reconstruction for Extraordinary Piezoelectricity

Active symmetry control - a central challenge in materials science, particularly in ferroelectrics - is achieved via mechanically assisted poling (MAP) guided by thermodynamics and phase - field modeling. This approach yields extraordinary piezoelectric coefficients (about 5,000 pC/N at 24 degC; 11,700 pC/N at 58 degC) together with about 65% optical transmittance in a classic relaxor ferroelectric, Pb(Mg1/3Nb2/3)O3-PbTiO3. Mechanical suppression of undesirable phases stabilizes a reconstructed symmetry with highly ordered domains, verified by multiple characterization techniques. The strategy is validated across several distinct ferroelectric systems. To demonstrate its practical utility, we fabricate a transparent dual-modal wearable sensor integrating continuous blood pressure monitoring via piezoelectricity with photoplethysmographic SpO2 detection, enabling high-fidelity physiological tracking. This work establishes mechanically assisted symmetry reconstruction as a pathway to multifunctional optoelectronic materials and compact wearable health technologies.

cond-mat.mtrl-sci↗

arXiv2026

Rigidity-Based Multi-Finger Coordination for Precise In-Hand Manipulation of Force-Sensitive Objects

Precise in-hand manipulation of force-sensitive objects typically requires judicious coordinated force planning as well as accurate contact force feedback and control. Unlike multi-arm platforms with gripper end effectors, multi-fingered hands rely solely on fingertip point contacts and are not able to apply pull forces, therefore poses a more challenging problem. Furthermore, calibrated torque sensors are lacking in most commercial dexterous hands, adding to the difficulty. To address these challenges, we propose a dual-layer framework for multi-finger coordination, enabling high-precision manipulation of force-sensitive objects through joint control without tactile feedback. This approach solves coordinated contact force planning by incorporating graph rigidity and force closure constraints. By employing a force-to-position mapping, the planned force trajectory is converted to a joint trajectory. We validate the framework on a custom dexterous hand, demonstrating the capability to manipulate fragile objects-including a soft yarn, a plastic cup, and a raw egg-with high precision and safety.

cs.RO↗

arXiv2026

Think Fast and Slow: Step-Level Cognitive Depth Adaptation for LLM Agents

Large language models (LLMs) are increasingly deployed as autonomous agents for multi-turn decision-making tasks. However, current agents typically rely on fixed cognitive patterns: non-thinking models generate immediate responses, while thinking models engage in deep reasoning uniformly. This rigidity is inefficient for long-horizon tasks, where cognitive demands vary significantly from step to step, with some requiring strategic planning and others only routine execution. In this paper, we introduce CogRouter, a framework that trains agents to dynamically adapt cognitive depth at each step. Grounded in ACT-R theory, we design four hierarchical cognitive levels ranging from instinctive responses to strategic planning. Our two-stage training approach includes Cognition-aware Supervised Fine-tuning (CoSFT) to instill stable level-specific patterns, and Cognition-aware Policy Optimization (CoPO) for step-level credit assignment via confidence-aware advantage reweighting. The key insight is that appropriate cognitive depth should maximize the confidence of the resulting action. Experiments on ALFWorld and ScienceWorld demonstrate that CogRouter achieves state-of-the-art performance with superior efficiency. With Qwen2.5-7B, it reaches an 82.3% success rate, outperforming GPT-4o (+40.3%), OpenAI-o3 (+18.3%), and GRPO (+14.0%), while using 62% fewer tokens.

cs.AI↗

arXiv2026

Coexistence of Antiferromagnetic Spin Fluctuations and Superconductivity in La2SmNi2O7 Thin Films

The interplay between magnetic fluctuations and superconductivity is fundamental for understanding unconventional high-temperature superconductors. In the recently discovered Ruddlesden-Popper phase nickelates-which achieve superconducting transition temperatures up to ~100 K-this connection has been theoretically predicted but experimentally unverified. Using compressively strained La2SmNi2O7 thin films, we report direct evidence for this interplay. We observe a characteristic 'Mexican hat'-shaped magnetoresistance, a signature of superconductivity coexisting with emergent antiferromagnetic (AFM) fluctuations. This distinctive feature arises from a competition between the magnetic field's suppression of AFM fluctuation-driven scattering and its suppression of superconducting fluctuations. We quantify these AFM fluctuations by a crossover field, B*, whose magnitude decreases with increasing temperature and vanishes near the superconducting onset-behavior contrasting sharply with that in high-Tc cuprates. Our results establish not merely coexistence, but also a direct and novel correlation between AFM instability and superconductivity in nickelates, offering crucial insight into their pairing mechanism.

cond-mat.supr-con↗

arXiv2026

SWE-SQL: Illuminating LLM Pathways to Solve User SQL Issues in Real-World Applications

Resolution of complex SQL issues persists as a significant bottleneck in real-world database applications. Current Large Language Models (LLMs), while adept at text-to-SQL translation, have not been rigorously evaluated on the more challenging task of debugging SQL issues. To address this gap, we introduce BIRD-CRITIC, a new SQL issue debugging benchmark comprising 530 PostgreSQL tasks (BIRD-CRITIC-PG) and 570 multi-dialect tasks (BIRD-CRITIC-Multi), distilled from authentic user issues and replayed within new environments to facilitate rigorous evaluation. Baseline evaluations underscore the task's complexity, with the leading reasoning model O3-Mini achieving only 38.87% success rate on BIRD-CRITIC-PG and 33.33% on BIRD-CRITIC-Multi. Meanwhile, advancing open-source models for database tasks is crucial for empowering local development while safeguarding data privacy. Therefore, we present Six-Gym (Sql-fIX-Gym), a training environment for elevating open-source model capabilities for SQL issue debugging. This environment leverages SQL-Rewind strategy, which automatically generates executable issue-solution datasets by reverse-engineering issues from verified SQLs. However, popular trajectory-based fine-tuning methods do not explore substantial supervisory signals. We further propose f-Plan Boosting, which extracts high-level debugging plans from SQL solutions, enabling teacher LLMs to produce 73.7% more successful trajectories for training. We integrate these components into an open-source agent, Bird-Fixer. Based on Qwen-2.5-Coder-14B, Bird-Fixer achieves 38.11% success rate on BIRD-CRITIC-PG and 29.65% on BIRD-CRITIC-Multi, surpassing leading proprietary models such as Claude-3.7-Sonnet and GPT-4.1, marking a significant step toward democratizing sophisticated SQL-debugging capabilities. The leaderboard and source code are available: https://bird-critic.github.io/

cs.DB↗

arXiv2026

Robust heterodimensional cycles in two-parameter unfolding of homoclinic tangencies

We establish a necessary and sufficient condition for the birth of heterodimensional cycles from a generic homoclinic tangency to a hyperbolic periodic orbit. We prove for $C^r$ ($r=3,\dots,\infty,ω$) dynamical systems on a manifold $\mathcal{M}$, with $\dim \mathcal{M}\geqslant 3$ for diffeomorphisms and with $\dim \mathcal{M}\geqslant 4$ for flows, that $C^1$-robust heterodimensional dynamics of coindex one appear in any generic two-parameter $C^r$ unfolding of a homoclinic tangency to a periodic orbit such that at least one central multiplier is not real and the central dynamics are not sectionally dissipative. The heterodimensional dynamics also involve a blender exhibiting $C^1$-robust homoclinic tangencies. As a corollary, any system with a homoclinic tangency of the class described above belongs to the $C^r$ closure of the $C^1$-open Newhouse domain.

math.DS↗

arXiv2026

A spectroscopically confirmed, strongly lensed, metal-poor Type II supernova at z = 5.13

Observing supernovae (SNe) in the early Universe (z > 3) provides a window into how both galaxies and individual stars have evolved over cosmic time, yet a detailed study of high-redshift stars and SNe has remained difficult due to their extreme distances and cosmological redshifting. To overcome the former, searches for gravitationally lensed sources allow for the discovery of magnified SNe that appear as multiple images - further providing the opportunity for efficient follow-up. Here we present the discovery of "SN Eos": a strongly lensed, multiply-imaged, SN II at a spectroscopic redshift of z = 5.133 +/- 0.001. SN Eos exploded in a Lyman-α emitting galaxy when the Universe was only ~1 billion years old, shortly after it reionized and became transparent to ultraviolet radiation. A year prior to our discovery in JWST data, archival HST imaging of SN Eos reveals rest-frame far ultraviolet (~1,300Å) emission, indicative of shock breakout or interaction with circumstellar material in the first few (rest-frame) days after explosion. The JWST spectroscopy of SN Eos, now the farthest spectroscopically confirmed SN ever discovered, shows that SN Eos's progenitor star likely formed in a metal-poor environment (<= 0.1 Z_{\odot}), providing the first direct evidence of massive star formation in the metal-poor, early Universe. SN Eos would not have been detectable without the extreme lensing magnification of the system, highlighting the potential of such discoveries to eventually place constraints on the faint end of the cosmic star-formation rate density in the very early Universe.

astro-ph.HE↗