arXiv subjects

Yan Li

Publications and source records attributed to Yan Li.

At least 19 recordsLinked to original sources

arXiv2026

Bayesian optimization enhanced magnetic field gradient detection based on spin-orbit coupled Bose-Einstein condensates

Magnetic field detection is of great significance in modern fields such as medical diagnostics, geological exploration, inertial navigation, and space physics. We have previously investigated the expansion dynamics of a spin-orbit-coupled Bose-Einstein condensate. Building on these results, here we propose a scheme for magnetic-field-gradient detection by establishing a quantitative relationship between the condensate's rotational deflection angle and the magnetic-field gradient. In this work, we present the parameter dependence of this scheme and perform a systematic optimization. We introduce a Bayesian optimization algorithm to automatically and efficiently search for the optimal combination of key parameters, thereby significantly enhancing the sensitivity of the magnetic field gradient measurement. The results show that the optimized gradiometer improves the sensitivity by a factor of five. This work not only provides a new approach for magnetic field gradient measurement but also serves as a reference for machine learning-empowered quantum precision measurement.

cond-mat.quant-gas↗

arXiv2026

Synchronization Mpemba Effect

We introduce the synchronization Mpemba effect, in which a less synchronized initial state overtakes a more synchronized one, and demonstrate it in dissipatively coupled quantum van der Pol oscillators. Within a graph-Laplacian framework, we show that for two oscillators, the mechanism is a competition between a slowly decaying collective background and a fast transverse transient, equivalently an exchange between relative-mode displacement and phase-fluctuation errors. The same decomposition extends to networks: the synchronization error splits into transverse Laplacian modes, yielding closed-form crossing times from eigenvalue separations and initial mode weights, and a sufficient screening criterion. With only these eigenvalues entering, the rule is topology independent: crossing times of different networks collapse onto a single spectral-gap line. Thus, quantum many-body dynamics can be designed and controlled at low dimension without constructing the exponentially large Liouvillian, opening a route to quantum synchronization as a resource.

quant-ph↗

arXiv2026

Multi-Partitioned Computing Quantum-Particle Approach: A Hybrid Quantum Framework for Fluid Flow

This study established a quantum-classical hybrid framework that integrates quantum computing paradigm with meshfree finite particle method. By harnessing quantum superposition and entanglement, it hybridized the critical computational kernels (termed as quantum finite particle method). A resource-efficient quantum computational strategy on multi-partitioned zones was proposed, which leverages a fixed small-scale quantum circuit as a fundamental processing unit to handle inner product for arbitrarily sized arrays. This approach employs iterative nesting of the quantum-core operation to accommodate varying input dimensions while maintaining hardware feasibility throughout. Motivated with developed quantum framework, the novel numerical discretization for hybrid quantum computational particle dynamics can be derived commonly and applied in fluid flows. Through a sequence of numerical experiments purposefully, the proposed numerical model was thoroughly validated and analyzed. Results demonstrate that integrating quantum computing to hybridize conventional linear combinations of particle dynamics serves as a novel computing paradigm. By further extending into the numerical investigation of viscoelastic, highly elastic, and purely elastic fluids under high Weissenberg number conditions, the applicability of simulation framework is broadened. Despite bottlenecks in quantum hardware and computational efficiency on this process, these advances offer critical insights for transitioning quantum-enhanced fluid simulation to practical engineering applications.

physics.flu-dyn↗

arXiv2026

BEND 2.0: Benchmarking the Capabilities Needed to Design DNA

Genomic foundation models are increasingly used to propose DNA sequences and edits. But successful design requires capabilities that current benchmarks do not measure. Here we introduce BEND 2.0 (BENchmark for DNA language models), which tests how well models predict the effects of DNA changes and whether these predictions can guide sequence design. Across eight tasks built on five experimental datasets, we find that model choice and how representations are accessed both affect performance. Linear readouts fitted to frozen representations can match or exceed models' native outputs, and functional post-training substantially improves the models' capabilities. Linear readouts transfer well across experimental assays and genes, suggesting that models capture functional patterns beyond the data used to fit individual predictors. At the same time, strong prediction of average effects can conceal weaker prediction of context-specific effects, and accurate prediction of combined mutation effects does not imply accurate prediction of their interactions. In design tasks, model predictions can reduce the number of candidates needed to find a desired response, but still leave substantial room for better design choices, particularly for context-specific and multi-cell objectives. Our benchmark reveals that genomic models now display signs of general utility but have yet to saturate on performance in even simple tasks.

q-bio.GN↗

arXiv2026

Symmetric Policy Optimization for Zero-Sum Markov Games

We propose a simple and symmetric method for policy optimization in discounted zero-sum Markov games. The method, termed symmetric policy iteration, operates by evaluating the joint policy and solving the corresponding regularized matrix game at each iteration. By employing a diminishing sequence of regularization strengths, we establish the global linear convergence of the proposed method to the optimal value function and the saddle point. In particular, the method does not require solving any standalone Markov decision process throughout policy optimization, and is symmetric in the sense that it stays invariant regardless of whether we are solving the min-max or the max-min version of the game. To the best of our knowledge, this appears to be the first simple and symmetric policy-based method attaining global convergence for discounted zero-sum Markov games.

math.OC↗

arXiv2026

An Embarrassingly Simple Data Selection Strategy for Million-Scale Multi-Modal Instruction-Tuning

Recent research demonstrates that pretrained large language models (LLMs) require minimal but carefully curated data for instruction-tuning. Building on LLMs, multi-modal models further contend with far larger data volumes, heterogeneous sources, and cross-modal alignment, amplifying both the necessity and the complexity of data selection. However, existing methods rely on deeply extracted embeddings, gradients or narrow quality scores that are time-consuming to compute and limited in expressiveness, frequently failing to surpass a simple random-sampling baseline. To overcome these limitations, we step back from complex embedding spaces and target the fundamental objective of instruction tuning: how and what knowledge should be elicited from the pretrained model and exposed to users. We leverage interaction \underline{S}tyles as a proxy for structural diversity, and decompose data quality into capability-specific \underline{S}cores, yielding an explicit yet \underline{R}ich (SSR) profile for each sample. We further cast the selection as submodular coverage maximization over the capability--style grid, providing an approximation guarantee. SSR consistently delivers superior performance across two distinct architecture stacks, LLaVA-1.5 (Vicuna + CLIP) and LLaVA-OneVision (Qwen2 + SigLIP), two large-scale data pools, six data budget settings, and 14 multi-modal benchmarks. To our knowledge, SSR is the first data selection method validated at million-scale, achieving 99.1% of full-data performance using only 30% of a 2.6M pool. Furthermore, the SSR profile models and selected subsets transfer training-free across data pools and model architectures with capability extensibility.

cs.CV↗

arXiv2026

Local regularity for solutions to singular and degenerate quasilinear parabolic equations

The main purpose of this paper is to prove local Hölder continuity for weak solutions to quasilinear parabolic equations of $p$-Laplace type under finite integrability assumptions on the forcing, covering the supercritical singular range and the entire degenerate regime. The key ingredient is an $L^\infty$-valued temporal Hölder estimate of the solution after any fixed positive waiting time, which is achieved by coupling a time-rescaling argument with De Giorgi iteration. We also provide counterexamples involving forcing terms with weaker spatial integrability to show that even bounded continuous solutions with compatible data may possess no positive temporal Hölder exponent. In particular, these examples establish the sharpness of the spatial integrability threshold in the degenerate case. Finally, a separate counterexample is constructed to demonstrate that the temporal Lipschitz endpoint can fail for bounded continuous forcing when it depends on time.

math.AP↗

arXiv2026

Patient-specific rubric-based evaluation of large language models across longitudinal cancer care

Large language models (LLMs) show promise in specialty medicine, yet existing evaluations largely focus on isolated clinical encounters rather than tracking how clinical responsibilities evolve across a longitudinal care pathway. Here we introduce NPC-RubricBench, a phase-aware benchmark for evaluating multistage clinical reasoning across the cancer care continuum, using nasopharyngeal carcinoma as a defined disease context. Curated from longitudinal records of 405 patients across two cancer centers, the benchmark comprises 941 patient-phase units across six sequential care phases and 17,688 patient-specific atomic criteria spanning five evaluation dimensions, with each assessment strictly restricted to decision-time clinical evidence. Evaluating 30 general-purpose and medical LLMs revealed that strong diagnostic performance does not carry over to downstream care. We evaluated 30 general-purpose and medical LLMs, showing that high diagnostic performance did not translate into improved downstream care. Instead, the dominant failure modes reversed across phases. Explicit decision errors dominated anatomical staging at 48.2%, whereas decision omissions dominated initial work-up and post-treatment follow-up at 33.4% and 30.4%, respectively. In the evaluated model set, medically adapted models showed higher omission rates than general-purpose models. Moreover, varying query granularity exposes a critical specification trade-off: broad open-ended requests degrade Decision Quality across all 30 models, while more detailed, checklist-like prompts raise Decision Quality but reduce Evidence Faithfulness and Safety Adherence. Furthermore, NPC-RubricBench demonstrates that static diagnostic benchmarks are insufficient to ensure consistent clinical reliability over time, highlighting the importance of phase-adaptive safeguards in multistage medical AI.

cs.AI↗

arXiv2026

Fast Instinct, Slow Reason: A Dual-Speed Agentic Framework for Self-Configuring Multi-Object Tracking in Satellite Video

Multi-object tracking in satellite video underpins persistent monitoring, yet offline-tuned trackers lack the onboard intelligence and efficiency to adapt within an observation of tens of seconds. The root is that adaptation still means relabelling and retraining the detector, which cannot run on board in time. Tuning the tracker configuration strongly controls accuracy without touching the network, yet search needs labelled or proxy evaluations, and a language agent costs seconds and tokens per decision. We propose a dual-speed agentic framework pairing a fast instinct compiled offline with a slow reason serving the mission. Its algorithm, Arbitration of Unlabelled Shortlists from Prior Experience, or AUSPEX, learns a covering shortlist and a label-free ranker from labelled source tasks and commits one configuration per sequence. Its system, a Cascaded Onboard Runtime with Verified Uplinked Skills, or CORVUS, cascades reflex, skill, and agent layers, so that a verified skill commits 2.5 s into each observation while a language agent serves natural- language missions. On SatMTB and SatVideoDT with three recent detectors, AUSPEX raises IDF1 by 2.84 to 2.94 points and MOTA by 3.21 to 3.29 points on average without target labels, balances accuracy and labels best among sixteen methods, and loses at most 0.28 IDF1 points at 1.25 tracking passes. CORVUS executes four missions with execution accuracy up to 0.99 and target recall at the perception ceiling, spends no token on the routine decision, runs on Jetson Xavier NX, and hosts fifteen language models from six families without altering any commitment, demonstrating its extensibility and generalization.

cs.CV↗

arXiv2026

Learning Native Reflection in Unified Models with Interleaved Reinforcement Learning

Unified multimodal models can both look at and render images, so in principle they can repair their own generations: diagnose what an image gets wrong, revise it, observe the result, and diagnose again. Whether a revision helps is known only after it is rendered, so the reflection text and the image generation must be learned jointly, over the whole loop. Supervised fine-tuning (SFT) on reflection trajectories gives a cold start but does not find the high-success repair paths, and naive RL that optimizes only the renderer or only one head leaves most of the gain untapped. We introduce UMM-Reflection, which applies reinforcement learning (RL) to complete reflection trajectories inside one unified model: sibling trajectories share one initial image, so the group-relative advantage compares reflection strategies, and one trajectory-level advantage updates both the reflection tokens and the flow-based revisions, avoiding the combinatorial blow-up of per-round credit assignment. Unlike single-round editing or pipelines with an external critic, credit flows across rounds and to both roles of the same model, and no verifier is needed at inference. On BAGEL, UMM-Reflection improves GenEval by 12.05 points over SFT, and the gains transfer to WISE (+10.97), OneIG-Bench (+3.48), and T2I-CompBench++ (+4.63), none of which is used in training.

cs.CV↗

arXiv2026

A Measurement Perspective on Benchmarking LLM-Generated Text Detectors using Item Response Theory

Benchmarking large language model (LLM)-generated text detectors commonly relies on aggregate detector-level metrics such as accuracy, F1 score, and AUC. Although useful for summarizing overall performance, these metrics provide limited insight into how benchmark composition shapes detector comparisons or which texts distinguish detectors with different response profiles. We propose a measurement-based benchmarking framework for jointly characterizing benchmark heterogeneity and detector behavior using multidimensional item response theory (IRT). By treating benchmark texts as measurement items and detectors as respondents, the framework estimates text-level discrimination and difficulty-location parameters alongside multidimensional detector abilities. We apply the framework to a benchmark of 32K human-written and machine-generated texts across four domains, evaluated by 14 detectors. The estimated IRT item-parameter space reveals distinct but overlapping human-machine distributions, with substantial variation across domains and text lengths. The human-machine label shows the strongest multivariate association with this space, while domain and text length further contribute to benchmark heterogeneity. Detector ability profiles reveal multidimensional and domain-dependent response patterns that are not fully reflected in aggregate performance rankings. As a practical application, clustering texts in the estimated IRT space yields a compact subset of 2,088 texts (6.5\% of the full 32K benchmark) that largely preserves the comparative detector-performance patterns observed on the full benchmark. Overall, the proposed framework provides a principled and interpretable methodology for benchmarking heterogeneous detection technologies while jointly characterizing the measurement properties of the texts used for their evaluation.

cs.CL↗

arXiv2026

Editing as Language: Teaching LLM Agents a Grammar for Native Context Editing

A synopsis in novels folds a story into a few lines; a function in code gathers many steps with a single execution. Context editing is native to language itself. In this paper, we propose the Native Context Editing (NCE) framework, which gives language model agents a grammar for natively editing their own context. By encoding context-editing operations in a grammar of dedicated special tokens, we teach the agents to invoke edits through the language itself. We train agents on iteratively constructed trajectory segments, with each teacher edit conditioned on the evolving context. This establishes context editing as an atomic and inherent capability of LLM agents, enabling them to dynamically modify their context as they generate responses over long-horizon reasoning and interaction. Compared with context-editing methods that learn to generate tool-use code within specific external harness environments, our approach integrates context editing as an internal mechanism of the model's native generation process, enabling better generalization across diverse real-world workflows. On four long-horizon agentic benchmarks (Terminal-Bench 2.0, SWE-bench Verified, SWE-bench Multilingual, and SWE-bench Pro), NCE improves the average task success rate over the base model by 5.6 percentage points for LongCat-Flash-Lite-Sparse and by 4.5 for Qwen3.6-35B-A3B, while reducing average token usage by 36% and 15%, respectively.

cs.AI↗

arXiv2026

Editing Images into Reusable Layers via Compositional Edit-Prior Distillation

High-quality image editing requires precise modification of target regions while preserving the content of unrelated regions. Existing editing models produce flattened outputs that limit precise local manipulation and layer reuse, while decomposition methods typically optimize layer recovery separately from instruction-driven editing. To this end, we introduce LayerEditor, a unified framework that couples instruction-guided image editing with controllable generation of reusable RGBA layers. At its core, In-Context RGB-Alpha Token Modeling jointly represents the edited canvas and an ordered sequence of RGB-Alpha layers within a single token sequence, enabling semantic edits and explicit control over individual layer content. To retain the pretrained model's editing capability during adaptation to layered outputs, we propose Compositional Edit-Prior Distillation, which transfers its editing prior to both the edited canvas and the image composed from the predicted layers. To support training and evaluation, we construct LayerEditor-100K, a large-scale paired image-editing dataset with ordered RGBA layer annotations, and introduce dedicated benchmarks for layered editing. Extensive experiments demonstrate that LayerEditor maintains strong editing fidelity while generating accurate RGBA layers, supporting both flexible layer-level manipulation and multi-layer decomposition.

cs.CV↗

arXiv2026

Interfacial Faraday instability driven by interspecies-interaction modulation in a phase-separated binary Bose-Einstein condensate

We investigate interfacial Faraday instabilities in a quasi-two-dimensional, phase-separated binary Bose-Einstein condensate driven by periodic modulation of the interspecies scattering length. In the strongly immiscible regime, the interspecies interaction energy is concentrated near the internal interface, linking the modulation to interfacial tension and the capillary restoring force. We quantify the first-order tension response through its logarithmic sensitivity and account for nonlinear tension variations at finite modulation amplitudes. A sharp-interface capillary model and Floquet analysis yield discrete azimuthal-mode frequencies and primary instability tongues that shift toward higher driving frequencies with increasing mode order. Gross-Pitaevskii simulations reveal star-shaped interface patterns with threefold to eightfold rotational symmetry. The eightfold mode exhibits subharmonic oscillations and approximately exponential growth of its period-averaged amplitude over a small-deformation interval. These results provide a route to frequency-selective excitation of interfacial capillary modes through interspecies-interaction control.

cond-mat.quant-gas↗

arXiv2026

Robust Reinforcement Learning for Dynamic Treatment Regimes under Model and Measurement Uncertainty

We develop a robust reinforcement learning (RL) framework for learning adaptive treatment policies that generalize across patients and remain reliable under errors in the measured biomarker and shifts in the disease dynamics. By modeling the patient as a hidden context, we address generalization and robustness in a unified manner through robust contextual Markov decision processes. Applying our framework to adaptive therapy for metastatic prostate cancer in a simulator calibrated to clinical data, we obtain generalizable and robust policies that substantially outperform the existing adaptive therapy method. In particular, we attain robustness with almost no loss of performance when the biomarker is accurate and the disease evolves as anticipated, and find that the resulting policy Pareto-dominates the existing method across a diverse array of scenarios. Our findings suggest that robust RL holds great promise as a principled methodology for learning reliable adaptive treatment policies.

cs.LG↗

arXiv2026

Moment-Video: Diagnosing Temporal Fidelity of Video MLLMs on Momentary Visual Events

Video multimodal large language models (MLLMs) have made rapid progress in general and long-form video understanding, yet their ability to capture and use brief, answer-critical visual evidence remains underexplored. Momentary actions or state transitions may determine an answer while occupying only a small fraction of the video. Performance on broad video understanding tasks alone does not establish whether models can reliably identify and use this temporally localized evidence. We introduce Moment-Video, a benchmark for evaluating temporal fidelity through momentary visual event understanding. Questions are constructed around visually observable events or event sequences, requiring models to notice, count, describe, or reason about transient evidence. The benchmark contains 1,000 human-verified video-QA pairs across seven domains and 25 subcategories, covering four tasks: Temporal Occurrence, Temporal Counting, Action Description, and Temporal Reasoning. We evaluate 34 proprietary and open-source MLLMs. Under the main evaluation settings, the best-performing model, Seed-2.0-Pro, achieves 39.6% overall accuracy, while most open-source models remain below 25%. Diagnostic analyses show that denser frame sampling improves some models, but substantial errors persist even when sampled frames cover all annotated necessary event intervals in an audited subset. Performance also declines on longer videos, suggesting additional temporal-localization challenges. By exposing these limitations, Moment-Video provides a diagnostic benchmark to guide the evaluation and development of video MLLMs that more reliably understand brief, answer-critical visual events.

cs.CV↗

arXiv2026

Excited-State Effects in Nucleon Sigma Terms from a Variational Analysis

We investigate excited-state effects in nucleon sigma-terms in lattice QCD using twisted-mass clover-improved fermion ensembles. We study correlation functions with nucleon and multihadron nucleon-$σ$-meson interpolating operators. The analysis is performed using three gauge ensembles, including one ensemble simulated with two degenerate light, a strange, and a charm quark with quark masses tuned to approximately their physical value. We solve the generalized eigenvalue problem that arises from the matrix of correlators and show that the three-point function for the light sigma-term obtained using the optimal linear combination of interpolating operators has substantially reduced time dependence without loss of statistical precision. We apply Laplace filtering and investigate residual excited-state contributions in the extraction of $σ_{πN}$. Using the same optimal linear combination of correlators as for $σ_{πN}$ for the strange sigma-term we also observe suppressed excited state contributions over an extended range of source--sink time separations. For the charm sigma-term, we observe negligible time dependence within the large statistical uncertainties, and these improvements show no significant effect.

hep-lat↗