arXiv subjects

Siqi Zhao

Publications and source records attributed to Siqi Zhao.

At least 19 recordsLinked to original sources

arXiv2026

ThinkPrior: Zero-Rollout Difficulty Priors for Cold-Start Prompt Selection in RLVR

In reinforcement learning with verifiable rewards (RLVR) trained with group relative policy optimization (GRPO), the KL-free reward-advantage term studied here depends on within-group reward variation. If all rollouts in a group are correct or all are wrong, their group-relative advantages are identically zero; these zero-advantage silent groups provide no reward-advantage gradient, yet uniform sampling spends 39% of a run's rollouts on them. History-based prompt selection must first spend target-policy rollouts to estimate difficulty, creating a cold start with rollout waste; ThinkPrior instead uses an external anchor in one offline pass to construct a zero-rollout difficulty prior before the first target-policy rollout. The verifier-scored anchor pass rate supplies an external-anchor initialization for a Beta posterior; ThinkPrior selects by expected learnability and then updates from training outcomes, changing neither the loss nor the optimizer. On Qwen2.5-Math-7B across sixteen seeds, ThinkPrior more than halves early silent groups and cuts wasted rollouts through step 30 by nearly a fifth, while we detect no difference in final accuracy. On this 250-prompt pool the fixed-budget result is a reallocation rather than a net saving. The measured ThinkPrior+DAPO composition reduces generated rollouts by 10.6% while both arms retain the same 3840-rollout update budget. The prior requires no target-policy rollout before the first selection, but the posterior thereafter uses target-policy outcomes.

cs.LG↗

arXiv2026

UniDiffFusion: A Unified Diffusion Framework for Multi-Task and Degradation-Robust Image Fusion

General image fusion aims to integrate complementary information from multiple source images, but existing methods often rely on task-specific models and struggle to maintain robust performance under diverse degradation conditions. In this paper, we propose UniDiffFusion, a unified diffusion framework for multi-task and degradation-robust image fusion. UniDiffFusion leverages the strong generative prior of a pretrained diffusion model to establish a shared fusion backbone across heterogeneous fusion tasks, while introducing task- and degradation-aware conditional adaptation to accommodate their distinct information-selection requirements. Specifically, we employ task prompt modulation to progressively adapt the shared diffusion representations to different fusion objectives, and develop a degradation prompt router to dynamically retrieve degradation-aware priors and restore corrupted source features before fusion. Furthermore, an application prompt bank is introduced to incorporate task-oriented semantic guidance for downstream applications, such as object detection and semantic segmentation, without altering the shared fusion and restoration pathways. The proposed framework is trained in a progressive manner to decouple fusion learning, degradation-aware restoration, and application-specific adaptation, thereby reducing interference among heterogeneous objectives. Extensive experiments on visible-infrared, multi-exposure, and multi-focus image fusion demonstrate that UniDiffFusion achieves superior fusion quality and robustness under both clean and degraded conditions. Moreover, UniDiffFusion consistently improves downstream detection and semantic segmentation performance, demonstrating its effectiveness as a unified diffusion framework for both perceptual fusion and task-oriented vision.

cs.CV↗

arXiv2026

A-PAIR: A Benchmark and Identity-Consistent Grounding Framework for Air-Ground Cross-View Referring Person Detection

Air-ground cross-view referring person detection is a necessary component in the language-to-perception-to-control chain of collective embodied intelligence, grounding a language command into the same physical target before ground and aerial agents can coordinate downstream actions. Existing referring expression comprehension and open-vocabulary grounding methods do not jointly account for cross-view identity consistency, making them insufficient for Air-Ground Cross-View Referring Person Detection (AGCV-RPD), which involves similar pedestrian distractors, weak aerial appearance cues, and cross-view identity consistency. To study this problem, we introduce Air-Ground Paired Identity-Aware Referring (A-PAIR), the first comprehensive AGCV-RPD benchmark, containing 22,137 cross-view referring samples. To construct A-PAIR efficiently, we propose Factorized Annotation and Referential Alignment (FARA), a semi-automatic annotation framework that generates factorized referring descriptions and identity-consistency supervision at reduced cost. We propose Identity-Consistent Referring Grounding (ICRG), a framework that combines factorized referential grounding, candidate-completeness supervision, and cross-view consistency calibration for joint air-ground pair selection. ICRG improves ground, aerial, and pair-level detection over strong baselines, increasing pair F1 from 16.65% to 22.28%. These results show that AGCV-RPD requires paired detection and identity-consistent reasoning.

cs.CV↗

arXiv2026

Towards a new paradigm of scientific discovery with socialized artificial intelligence

Scientific discovery has advanced through successive transformations in the organization of knowledge. Observation and experimentation established the empirical foundations of science. Theory made it possible to derive general principles from particular phenomena. Computation extended inquiry into systems beyond direct observation, while data-intensive methods opened new spaces of pattern and prediction. Science now confronts a different frontier. The central challenge is no longer simply to produce more information, but to organize expanding knowledge, reasoning, and evidence into a coherent process of discovery. Here, we introduce Bridging Literature, Agents, and Zero-gap Experimentation (BLAZE), a paradigm of socialized scientific intelligence. BLAZE conceives AI not as an assistant for isolated research tasks, but as an organizational infrastructure for scientific discovery. It connects persistent knowledge, collective reasoning, empirical validation, and human judgment within a continuous research lifecycle, transforming fragmented activities into a cumulative process of inquiry, criticism, and revision. The central premise of BLAZE is that scientific intelligence does not arise from computation alone. It emerges from the sustained interaction among knowledge, hypotheses, experiments, and collective verification. By organizing humans and machines within a shared scientific process, BLAZE makes discovery more traceable, reproducible, and cumulative while preserving human creativity, judgment, and responsibility. Socialized scientific intelligence may provide a foundation for the next era of science. Its purpose is not to replace human discovery, but to extend the scale, depth, and continuity of collective scientific inquiry.

cs.AI↗

arXiv2026

Pitch-Angle Scattering of Cosmic Rays: Confronting Theory with Observations

Cosmic ray (CR) propagation is controlled by scattering in turbulent magnetic fields in space. In general, diffusive propagation is governed by pitch-angle diffusion in phase space. In this study, pitch-angle diffusion in the local interstellar medium (LISM) deduced from the analysis of the CR small-scale anisotropy data from the Tibet AS$γ$ experiment is compared with theoretical predictions. While it is difficult to reconcile the inferred LISM pitch angle diffusion coefficient with conventional theoretical results of particle scattering by Alfvénic turbulence, we find remarkable agreement with the predictions of particle scattering by quasi-slab fast modes whose properties are shaped by damping in the warm ionized medium. These findings offer direct evidence that CR scattering is predominantly governed by fast-mode turbulence. Furthermore, the comparison between experimental and theoretical results imposes strong constraints on plasma and magnetic field parameters within the local bubble, indicating that the LISM is in a low $β$ condition. The turbulence in the LISM should be compressible with an amplitude in the range of $0.3\lesssim δB/B_0 <1$.

astro-ph.HE↗

arXiv2026

Spatiotemporal Properties of Compressible Magnetohydrodynamic Turbulence from Space Plasma

Previous studies have established that a weak-to-strong transition occurs in Alfvenic magnetohydrodynamic (MHD) turbulence as energy cascades from large to small scales. However, the spatiotemporal (frequency-wavenumber) properties of compressible MHD turbulence involving all eigenmodes, which encode the strength of nonlinear interactions, remain difficult to characterize observationally. Consequently, whether a similar weak-to-strong transition occurs in compressible turbulence remains elusive. Using a novel multi-spacecraft, polarization-based mode-decomposition technique with measurements from the Cluster spacecraft in Earth's magnetosheath, we obtain spatiotemporal power spectra of all MHD eigenmodes and present the first quantitative assessment of nonlinear frequency broadening. Our results show that slow modes exhibit a weak-to-strong transition, evolving from wave-like peaks to frequency-broadened spectra as nonlinearity increases, whereas fast modes remain weakly turbulent with narrow peaks near their eigenfrequencies. Both Alfvenic and compressible fluctuations contribute significantly to low-frequency, large-scale quasi-two-dimensional structures. These findings provide a comprehensive observational characterization of compressible turbulence across mode composition, spatiotemporal scales, and weak-strong turbulence regimes, with implications for energetic particle transport, turbulent dynamos, plasma heating, and solar wind-magnetosphere coupling.

physics.plasm-ph↗

arXiv2026

A Method for Imaging Interplanetary Magnetic Field Strength and Orientation

Measurements of interplanetary magnetic fields have long relied on spacecraft measurements, which provide only in-situ sampling and therefore cannot capture the global magnetic structure. Faraday rotation of radio signals extends in-situ measurements to line-of-sight measurements, but it still depends on the number and spatial distribution of available radio sources. The Zeeman effect offers another route to remote sensing of magnetic fields, but it is generally too weak to diagnose the weak interplanetary magnetic fields. Here, we present a remote-sensing method to constrain weak magnetic field strength and orientation using spectral-line polarization induced by ground-state alignment (GSA) and Hanle effect, with collisional effects taken into account. This method is sensitive to weak magnetic fields in environments ranging from the high solar atmosphere and solar wind to the outer heliosphere, and we identify suitable spectral lines for different targets. We further perform forward modeling of Mercury's magnetosphere to demonstrate the feasibility of this imaging method. Spectral-polarization imaging therefore provides a new way toward remote imaging of dynamic heliospheric magnetic structures.

physics.space-ph↗

arXiv2026

Compressible Turbulence as a Source of Particle Beams and Ion Bernstein Waves in Collisionless Plasmas

Unraveling the origin of proton beams and ion Bernstein waves is important to understanding kinetic dissipation in the solar wind. Here we focus on their generation mechanisms, rather than their well-studied roles in instabilities and particle heating. We investigate their formation in collisionless plasmas using high-resolution particle-in-cell simulations of compressible turbulence. At magnetohydrodynamic (MHD) scales, compressive fluctuations are damped via transit-time damping (TTD), naturally producing suprathermal electrons and proton beams. At sub-ion scales, quasi-perpendicular fast modes excite multiple branches of ion Bernstein waves, whose properties agree with predictions from the plasma dispersion relation solver. Under solar wind conditions, TTD remains efficient and provides a natural explanation for the super-Alfvénic proton beams measured in situ. Our results demonstrate that compressive fluctuations play a central role in driving cross-scale energy transfer and kinetic dissipation in collisionless plasma turbulence.

physics.plasm-ph↗

arXiv2026

Solar Wind Reflected Ion Properties at Earth's Bow Shock: Dependence on Upstream Conditions and Shock Geometry

Solar wind ion reflection at collisionless shocks regulates foreshock plasma dynamics, yet the quantitative dependence of reflected ion properties on upstream and shock-related parameters remains unclear, causing difficulties in predicting foreshock disturbances. We present a statistical study of solar wind reflected ions near the Earth's bow shock using THEMIS observations from 59 well-defined shock crossings between 2016 and 2019. Reflected ion moments are derived after removal of the solar wind core and compared with upstream and shock parameters. The reflection ratio decreases with increasing angle between interplanetary magnetic field and shock normal, and increases with magnetic compression ratio, indicating that shock geometry and magnetic compression primarily regulate ion reflection. Reflected ion energies deviate from individual idealized reflection models: the adiabatic model overestimates total ion energy, whereas the specular model captures the perpendicular component. A combined adiabatic-specular representation improves the linear energy correspondence, and model-observation agreement increases under quasi-perpendicular shock conditions for all comparisons. Reflected ion temperature correlates with upstream magnetic field strength and solar wind temperature, and shows a substantially stronger dependence on magnetic field fluctuation energy. Overall, reflected ion properties are primarily governed by upstream conditions and shock structure, with magnetic field fluctuations contributing to ion thermalization, providing observational constraints on ion reflection and heating at Earth's bow shock.

physics.space-ph↗

arXiv2026

Automatic In-Domain Exemplar Construction and LLM-Based Refinement of Multi-LLM Expansions for Query Expansion

Query expansion with large language models is promising but often relies on hand-crafted prompts, manually chosen exemplars, or a single LLM, making it non-scalable and sensitive to domain shift. We present an automated, domain-adaptive QE framework that builds in-domain exemplar pools by harvesting pseudo-relevant passages using a BM25-MonoT5 pipeline. A training-free cluster-based strategy selects diverse demonstrations, yielding strong and stable in-context QE without supervision. To further exploit model complementarity, we introduce a two-LLM ensemble in which two heterogeneous LLMs independently generate expansions and a refinement LLM consolidates them into one coherent expansion. Across TREC DL20, DBPedia, and SciFact, the refined ensemble delivers consistent and statistically significant gains over BM25, Rocchio, zero-shot, and fixed few-shot baselines. The framework offers a reproducible testbed for exemplar selection and multi-LLM generation, and a practical, label-free solution for real-world QE.

cs.IR↗

arXiv2026

Neural Nonlinear Shrinkage of Covariance Matrices for Minimum Variance Portfolio Optimization

This paper introduces a neural network-based nonlinear shrinkage estimator of covariance matrices for the purpose of minimum variance portfolio optimization. It is a hybrid approach that integrates statistical estimation with machine learning. Starting from the Ledoit-Wolf (LW) shrinkage estimator, we decompose the LW covariance matrix into its eigenvalues and eigenvectors, and apply a lightweight transformer-based neural network to learn a nonlinear eigenvalue shrinkage function. Trained with portfolio risk as the loss function, the resulting precision matrix (the inverse covariance matrix) estimator directly targets portfolio risk minimization. By conditioning on the sample-to-dimension ratio, the approach remains scalable across different sample sizes and asset universes. Empirical results on stock daily returns from Standard & Poor's 500 Index (S&P500) demonstrate that the proposed method consistently achieves lower out-of-sample realized risk than benchmark approaches. This highlights the promise of integrating structural statistical models with data-driven learning.

cs.LG↗

arXiv2025

Mode Composition Shapes Magnetic Anisotropy in Solar Wind Turbulence

Turbulence is a ubiquitous process that transfers energy across many spatial and temporal scales, thereby influencing particle transport and heating. Recent progress has improved our understanding of the anisotropy of turbulence with respect to the mean magnetic field; however, its exact form and implications for magnetic topology and energy transfer remain unclear. In this study, we investigate the nature of magnetic anisotropy in compressible magnetohydrodynamic (MHD) turbulence within low-$β$ solar wind using measurements from the Cluster spacecraft. By decomposing small-amplitude fluctuations into Alfvén and compressible modes, we reveal that magnetic anisotropy is largely mode dependent: Alfvenic fluctuations are broadly distributed in propagation angle, whereas compressible fluctuations are concentrated near the quasi-parallel (slab) direction, a feature closely linked to collisionless damping of compressible modes. For $β\rightarrow0$, compressible modes become dominant within the slab component at smaller scales. These findings advance our understanding of magnetic anisotropy in solar wind turbulence and offer a new perspective on the three-dimensional turbulence cascade, with broad implications for particle transport, acceleration, and magnetic reconnection.

astro-ph.SR↗

arXiv2025

A Survey of Long-Document Retrieval in the PLM and LLM Era

The proliferation of long-form documents presents a fundamental challenge to information retrieval (IR), as their length, dispersed evidence, and complex structures demand specialized methods beyond standard passage-level techniques. This survey provides the first comprehensive treatment of long-document retrieval (LDR), consolidating methods, challenges, and applications across three major eras. We systematize the evolution from classical lexical and early neural models to modern pre-trained (PLM) and large language models (LLMs), covering key paradigms like passage aggregation, hierarchical encoding, efficient attention, and the latest LLM-driven re-ranking and retrieval techniques. Beyond the models, we review domain-specific applications, specialized evaluation resources, and outline critical open challenges such as efficiency trade-offs, multimodal alignment, and faithfulness. This survey aims to provide both a consolidated reference and a forward-looking agenda for advancing long-document retrieval in the era of foundation models.

cs.IR↗

arXiv2025

Energy Cascade and Damping in Fast-Mode Compressible Turbulence

Compressible turbulence governs energy transfer across scales in space and astrophysical systems. Capturing both the turbulence cascade and damping is therefore crucial for models of energy conversion, plasma heating, and particle transport in diverse plasma environments, but remains challenging. Progress is constrained by two unresolved fundamental questions: the persistence of the turbulence cascade in the presence of shocks and discontinuities, and the validity of classical wave theories under strong nonlinearity. In particular, it remains unclear whether meaningful cascade dynamics can be defined in compressible turbulence with phase steepening, and whether frameworks developed for monochromatic waves remain applicable to complex, broadband fluctuations. Using large-scale, high-resolution kinetic simulations, we analyze turbulence-particle interactions, which are beyond the capability of standard magnetohydrodynamic (MHD) simulations. We show that compressible turbulence damping at MHD scales in quantitative agreement with transit-time damping theory, even in fully developed nonlinear states. Moreover, the cascade persists despite the generation of shocks and discontinuities due to phase steepening, revealing a surprising robustness of cross-scale energy transfer under extreme conditions. We further provide the spectral expression of compressible turbulence. These results close a long-standing gap in the physics of compressible turbulence and establish a robust foundation for turbulence modeling from the heliosphere to galaxies.

astro-ph.SR↗

arXiv2025

Decoding Polyphenol-Protein Interactions with Deep Learning: From Molecular Mechanisms to Food Applications

Polyphenols and proteins are essential biomolecules that influence food functionality and, by extension, human health. Their interactions -- hereafter referred to as PhPIs (polyphenol-protein interactions) -- affect key processes such as nutrient bioavailability, antioxidant activity, and therapeutic efficacy. However, these interactions remain challenging due to the structural diversity of polyphenols and the dynamic nature of protein binding. Traditional experimental techniques like nuclear magnetic resonance (NMR) and mass spectrometry (MS), along with computational tools such as molecular docking and molecular dynamics (MD), have offered important insights but face constraints in scalability, throughput, and reproducibility. This review explores how deep learning (DL) is reshaping the study of PhPIs by enabling efficient prediction of binding sites, interaction affinities, and MD using high-dimensional bio- and chem-informatics data. While DL enhances prediction accuracy and reduces experimental redundancy, its effectiveness remains limited by data availability, quality, and representativeness, particularly in the context of natural products. We critically assess current DL frameworks for PhPIs analysis and outline future directions, including multimodal data integration, improved model generalizability, and development of domain-specific benchmark datasets. This synthesis offers guidance for researchers aiming to apply DL in unraveling structure-function relationships of polyphenols, accelerating discovery in nutritional science and therapeutic development.

q-bio.BM↗

arXiv2025

Guided Generation for Developable Antibodies

Therapeutic antibodies require not only high-affinity target engagement, but also favorable manufacturability, stability, and safety profiles for clinical effectiveness. These properties are collectively called `developability'. To enable a computational framework for optimizing antibody sequences for favorable developability, we introduce a guided discrete diffusion model trained on natural paired heavy- and light-chain sequences from the Observed Antibody Space (OAS) and quantitative developability measurements for 246 clinical-stage antibodies. To steer generation toward biophysically viable candidates, we integrate a Soft Value-based Decoding in Diffusion (SVDD) Module that biases sampling without compromising naturalness. In unconstrained sampling, our model reproduces global features of both the natural repertoire and approved therapeutics, and under SVDD guidance we achieve significant enrichment in predicted developability scores over unguided baselines. When combined with high-throughput developability assays, this framework enables an iterative, ML-driven pipeline for designing antibodies that satisfy binding and biophysical criteria in tandem.

cs.LG↗

arXiv2025

Observations of Turbulence and Particle Transport at Interplanetary Shocks: Transition of Transport Regimes

The transport of energetic particles is intimately related to the properties of plasma turbulence, a ubiquitous dynamic process that transfers energy across a broad range of spatial and temporal scales. However, the mechanisms governing the interactions between plasma turbulence and energetic particles remain incompletely understood. Here we present comprehensive observations from the upstream region of a quasi-perpendicular interplanetary (IP) shock on 2004 January 22, using data from four Cluster spacecraft to investigate the interplay between turbulence dynamics and energetic particle transport. Our observations reveal a transition in energetic proton fluxes from exponential to power-law decay with increasing distance from the IP shock. This result provides possible observational evidence of a shift in transport behavior from normal diffusion to superdiffusion. This transition correlates with an increase in the time ratio from $τ_s/τ_{c}<1$ to $τ_s/τ_{c}\gg1$, where $τ_s$ is the proton isotropization time, and $τ_{c}$ is the turbulence correlation time. Additionally, the frequency-wavenumber distributions of magnetic energy in the power-law decay zone indicate that energetic particles excite linear Alfvén-like harmonic waves through gyroresonance, thereby modulating the original turbulence structure. These findings provide valuable insights for future studies on the propagation and acceleration of energetic particles in turbulent astrophysical and space plasma systems.

physics.plasm-ph↗

arXiv2024

A Systematic Review of Machine Learning Methods for Multimodal EEG Data in Clinical Application

Machine learning (ML) and deep learning (DL) techniques have been widely applied to analyze electroencephalography (EEG) signals for disease diagnosis and brain-computer interfaces (BCI). The integration of multimodal data has been shown to enhance the accuracy of ML and DL models. Combining EEG with other modalities can improve clinical decision-making by addressing complex tasks in clinical populations. This systematic literature review explores the use of multimodal EEG data in ML and DL models for clinical applications. A comprehensive search was conducted across PubMed, Web of Science, and Google Scholar, yielding 16 relevant studies after three rounds of filtering. These studies demonstrate the application of multimodal EEG data in addressing clinical challenges, including neuropsychiatric disorders, neurological conditions (e.g., seizure detection), neurodevelopmental disorders (e.g., autism spectrum disorder), and sleep stage classification. Data fusion occurred at three levels: signal, feature, and decision levels. The most commonly used ML models were support vector machines (SVM) and decision trees. Notably, 11 out of the 16 studies reported improvements in model accuracy with multimodal EEG data. This review highlights the potential of multimodal EEG-based ML models in enhancing clinical diagnostics and problem-solving.

eess.SP↗