arXiv subjects

Chang Zhou

Publications and source records attributed to Chang Zhou.

At least 19 recordsLinked to original sources

arXiv2026

TradingMoE: Routing the Right Experts in Evolving Markets

Large language models (LLMs) have shown strong potential for financial analysis and trading, but direct trading remains challenging because the predictive capabilities required can vary across assets, decision fields, and market conditions. Existing LLM-based trading systems either coordinate human-defined external experts or adopt conventional internal Mixture-of-Experts (MoE) routers that do not directly evaluate how individual experts contribute to trading decisions. Moreover, these routers receive no direct signal indicating when an inactive expert has become more suitable as market conditions change. We find that native router scores poorly reflect how much individual experts improve trading decisions, frequently leaving better alternatives unselected. We further reveal that token-specific expert usefulness exhibits a compact low-dimensional structure. Based on these findings, we propose TradingMoE, a trading-oriented sparse MoE that augments a frozen dense LLM with lightweight residual experts. We introduce a Query-Key router that represents the expertise required by each token under the current market context as a low-dimensional query and matches it with learnable expert keys. We further propose a sparse expert selection update mechanism that samples a few inactive experts during training and estimates whether they should replace the weakest expert in the current Top-k route. This mechanism enables the router to update expert selection as market conditions change while preserving sparse computation. Experiments against 22 baselines on stock and cryptocurrency markets show that TradingMoE improves cumulative return over the best-performing baselines by 30.89% and 30.7%, respectively. Rolling paper-trading experiments further demonstrate that its advantage persists under forward-only deployment.

cs.LG↗

arXiv2026

Infrared Echoes of Precessing Tidal Disruption Events

A tidal disruption event (TDE) occurs when a star is torn apart by a supermassive black hole. The resulting UV/optical flare irradiates parsec-scale dust, producing delayed mid-infrared echoes that persist for years. These echoes provide unique calorimetric probes of the total radiated energy and dust geometry. Existing models usually assume static axisymmetric illumination patterns. However, the TDE accretion disk is likely misaligned and undergoes relativistic precession. In this work, we present a theoretical framework for infrared dust echoes from a precessing TDE disk. The precession will lead to highly variable infrared light curves, which can be revealed by high-cadence observations. The overall profile of the infrared light curves shows double-peaked to single-peaked pattern transitions as a result of the changes in the viewing angle or precession angle. The results indicate that infrared echoes are dynamic tracers of the evolving lighting patterns of the central engine.

astro-ph.HE↗

arXiv2026

Simulations of interaction between outflow and surrounding broken power-law circumnuclear medium: implications for different radio light curves of TDEs

The complex radio light curves of tidal disruption events (TDEs) challenge our understanding of the properties of both the outflows and the circumnuclear medium (CNM) surrounding supermassive black holes. In this work, we explore outflow-CNM interactions across a broad parameter space using three-dimensional hydrodynamic simulations, adopting a broken power-law CNM density profile with a transition near the Bondi radius. The outflow-CNM interaction inside Bondi radius produces an early radio flare ($\lesssim 2$ yr) once the emitting region becomes optically thin. A second radio rebrightening can appear a few years later if the outflow decelerates beyond Bondi radius. We also find that either a very dense inner CNM, which causes rapid deceleration, or a rarefied outer CNM suppresses the late rebrightening that will produces a single early-peaked flare. In contrast, a rarefied CNM inside the Bondi radius suppresses the early flare and yields a single late-peaked event. For the case of very dense CNM at large radii, the interaction will trigger a sharp late-time rise as observed in some TDEs. We further explore the interaction of a relativistic jet with a broken power-law CNM, which can reproduce the characteristic light curves as observed in jetted TDEs without invoking complex jet structure.

astro-ph.HE↗

arXiv2026

GraphReAct: Reasoning and Acting for Multi-step Graph Inference

Reasoning-acting frameworks enhance large language models (LLMs) by interleaving reasoning with actions for dynamic information acquisition. However, extending this paradigm to graph learning remains underexplored. Graph data is inherently structured, with information distributed across nodes and edges and encoded through both topology and latent representations. As a result, effective reasoning over graphs requires not only retrieving informative evidence from the graph, but also progressively refining the accumulated context during multi-step inference. In this work, we propose GraphReAct, a graph reasoning-acting framework that enables step-by-step inference over graph-structured data. Specifically, we design a graph-based action space with two complementary retrieval actions: topological retrieval, which captures local structural dependencies, and semantic retrieval, which accesses non-local but relevant evidence in the representation space. These actions dynamically expand the reasoning context. To further support multi-step reasoning, we introduce another type of action, context refinement, which distills and reorganizes accumulated information into a compact representation. By interleaving reasoning with both retrieval and refinement actions, our framework enables a progressive transition from context expansion to compression. Extensive experiments on six benchmark datasets demonstrate that GraphReAct consistently outperforms state-of-the-art methods, validating the effectiveness of reasoning-acting for graph learning.

cs.AI↗

arXiv2026

Multi-wavelength study of EP250416a / GRB 250416C: An Optically Dark Long GRB with a Late Jet Break

We present multi-wavelength study of the $\gamma$/X-ray transient EP250416a (also designated GRB 250416C), triggered by the Einstein Probe (EP) Wide-field X-ray Telescope and also by SVOM and Konus-Wind. Observations spanning the gamma-ray, X-ray, and optical bands facilitated detailed analysis of the burst's prompt emission, afterglow evolution, and physical origin. EP250416a exhibits a burst duration of 30 s in X-ray and 17.7 s in gamma-rays, with joint spectral fitting of 0.5-5000 keV data gives $E\rm_{peak}=342_{-232}^{+90}$ keV. Optical spectroscopy of the afterglow, acquired with the Gemini Multi-Object Spectrograph (GMOS) on Gemini South, yielded a redshift of $z=0.963$. Accounting for the measured redshift, the isotropic energies are $E\rm_{X,iso}=2.7_{-0.5}^{+0.9}\times10^{50}$ erg and $E\rm_{\gamma,iso}=7.34_{-2.1}^{+5.1}\times10^{51}$ erg, aligning with the Amati relation for long GRBs. The fluence ratio $\rm S(25-50~keV)/S(50-100~keV)=0.78_{-0.15}^{+0.1}$ classifies EP250416a as an X-ray rich (XRR) GRB. The X-ray afterglow shows an initial shallow decay ($\alpha \approx -0.5$) transitioning to a canonical decay phase ($\alpha \approx -1$), with a very late jet break at $t\sim 1.5\times 10^6$ s, corresponding to a jet half-opening angle of $\theta _j=10.6_{-1.8}^{+1.9}$ degrees. EP250416a is optically dark, as it shows only a faint $r$-band detection ($r=24.16$ mag) from Gemini South-GMOS and a low optical-to-X-ray spectral index $\beta_{\rm OX} = 0.3$. This may be attributed to significant host-galaxy extinction, with a required $A_V^{\text{host}}=5.5\ \text{mag}$ derived from the extinction curve model.

astro-ph.HE↗

arXiv2026

A Benchmark and Multi-Agent System for Instruction-driven Cinematic Video Compilation

The surging demand for adapting long-form cinematic content into short videos has motivated the need for versatile automatic video compilation systems. However, existing compilation methods are limited to predefined tasks, and the community lacks a comprehensive benchmark to evaluate the cinematic compilation. To address this, we introduce CineBench, the first benchmark for instruction-driven cinematic video compilation, featuring diverse user instructions and high-quality ground-truth compilations annotated by professional editors. To overcome contextual collapse and temporal fragmentation, we present CineAgents, a multi-agent system that reformulates cinematic video compilation into ``design-and-compose'' paradigm. CineAgents performs script reverse-engineering to construct a hierarchical narrative memory to provide multi-level context and employs an iterative narrative planning process that refines a creative blueprint into a final compiled script. Extensive experiments demonstrate that CineAgents significantly outperforms existing methods, generating compilations with superior narrative coherence and logical coherence.

cs.CV↗

arXiv2026

Investigating the Circumnuclear Medium of Tidal Disruption Events with Radio Observations

Tidal disruption events (TDEs) are unique tools for investigating quiescent supermassive black hole (SMBH), accretion physics, and circumnuclear medium (CNM) environments. The CNM density profile is of great astrophysical significance, since it provides key diagnostics for the accretion history of dormant SMBH. TDEs can launch outflows that produce radio emission when propagating into the CNM. The closure relation (CR), i.e., the relation between the temporal indices and the spectral indices, are therefore monitoring the CNM density profile. In this work, we first collect 53 TDEs with radio observations to date. We then obtain the predicted CR for arbitrary CNM and different dynamical phases of the outflow, and apply to the radio TDE sample. We constrain the CNM density profile for 26 radio TDEs with good data quality. The results are generally consistent with those estimated with equipatition method, suggesting that CR analysis is efficient in the study of CNM profile for a quiescent SMBH.

astro-ph.HE↗

arXiv2026

Crab$^{+}$: A Scalable and Unified Audio-Visual Scene Understanding Model with Explicit Cooperation

Developing Audio-Visual Large Language Models (AV-LLMs) for unified scene understanding is pivotal in multimodal intelligence. While instruction tuning enables pre-trained models with multi-task abilities, we observe that conventional multi-task unification methods often suffer from severe negative transfer, where nearly 55% of tasks degrade compared to single-task training. We attribute this phenomenon to audio-visual task heterogeneity, characterized by disparate task granularity and divergent capability demands, which lead to negative interference under joint training. To tackle this, we present Crab$^{+}$, a scalable and unified audio-visual scene understanding model that addresses task heterogeneity through explicit cooperation from both data and model perspectives. On the data side, we introduce AV-UIE v2, a comprehensive Audio-Visual Unified Instruction-tuning dataset with Explicit reasoning processes. It contains approximately 222K samples spanning 17 datasets and 7 tasks, enabling the model to capture cross-task relationships at different levels of granularity. On the model side, we design a unified interface to align heterogeneous task formulations, and propose Interaction-aware LoRA (I-LoRA), which explicitly models inter-task relationships via dynamic routing to coordinate distinct audio-visual interaction patterns, mitigating parameter interference. Extensive experiments show Crab$^{+}$ covers broader tasks than existing unified models while outperforming specialized models on various benchmarks. We successfully reverse the negative transfer trend, achieving positive transfer where multi-task learning surpasses single-task baselines in nearly 88% of tasks. These results hold across diverse AV-LLM paradigms and are validated through in-depth visualization, positioning Crab$^{+}$ as a robust step towards holistic audio-visual scene understanding.

cs.CV↗

arXiv2026

Evaluating Progress in Graph Foundation Models: A Comprehensive Benchmark and New Insights

Graph foundation models (GFM) aim to acquire transferable knowledge by pre-training on diverse graphs, which can be adapted to various downstream tasks. However, domain shift in graphs is inherently two-dimensional: graphs differ not only in what they describe (topic domains) but also in how they are represented (format domains). Most existing GFM benchmarks vary only topic domains, thereby obscuring how knowledge transfers across both dimensions. We present a new benchmark that jointly evaluates topic and format gaps across the full GFM pipeline, including multi-domain self-supervised pre-training and few-shot downstream adaptation, and provides a timely evaluation of recent GFMs in the rapidly evolving landscape. Our protocol enables controlled assessment in four settings: (i) pre-training on diverse topics and formats, while adapting to unseen downstream datasets; (ii) same pre-training as in (i), while adapting to seen datasets; (iii) pre-training on a single topic domain, while adapting to other topics; (iv) pre-training on a base format, while adapting to other formats. This two-axis evaluation disentangles semantic generalization from robustness to representational shifts. We conduct extensive evaluations of eight state-of-the-art GFMs on 33 datasets spanning seven topic domains and six format domains, surfacing new empirical observations and practical insights for future research. Codes/data are available at https://github.com/smufang/GFMBenchmark.

cs.CL↗

arXiv2026

APPO: Attention-guided Perception Policy Optimization for Video Reasoning

Complex video reasoning, actually, relies excessively on fine-grained perception rather than on expert (e.g., Ph.D, Science)-level reasoning. Through extensive empirical observation, we have recognized the critical impact of perception. In particular, when perception ability is almost fixed, enhancing reasoning from Qwen3-8B to OpenAI-o3 yields only 0.7% performance improvement. Conversely, even minimal change in perception model scale (from 7B to 32B) boosts performance by 1.4%, indicating enhancing perception, rather than reasoning, is more critical to improve performance. Therefore, exploring how to enhance perception ability through reasoning without the need for expensive fine-grained annotation information is worthwhile. To achieve this goal, we specially propose APPO, the Attention-guided Perception Policy Optimization algorithm that leverages token-level dense rewards to improve model's fine-grained perception. The core idea behind APPO is to optimize those tokens from different responses that primarily focus on the same crucial video frame (called intra-group perception tokens). Experimental results on diverse video benchmarks and models with different scales (3/7B) demonstrate APPO consistently outperforms GRPO and DAPO (0.5%~4%). We hope our work provides a promising approach to effectively enhance model's perception abilities through reasoning in a low-cost manner, serving diverse scenarios and demands.

cs.CV↗

arXiv2026

Minutes-long soft X-ray prompt emission from a compact object merger

Compact object mergers are multi-messenger sources and known progenitors of some gamma-ray bursts, bright flashes of high-energy radiation powered by a central engine, either an accreting black hole or a neutron star. Our understanding of these events has so far been shaped primarily by observations in the gamma-ray band, leaving their prompt phase poorly constrained at lower energies. A long-lasting ($\approx$100 s) engine-driven X-ray emission was discussed to explain rapidly fading X-ray afterglows following several ($\approx$30%) bursts of short ($\lesssim$2 s) duration. However, this prompt X-ray component was not directly observed and past candidates were not confirmed. Here we report the discovery of EP250704a containing a minutes-long ($\sim$560 s) flash of soft (0.5--4 keV) X-rays immediately following the short ($\sim$0.4 s) GRB 250704B. The variability and spectral shape of this emission are inconsistent with the canonical picture of a hard, accretion-powered spike followed by a standard external-shock afterglow. Instead, the long-soft bump points to a distinct phase of prompt emission in X-rays, which would not have been detected without the soft X-ray coverage of Einstein Probe. The detection of a prompt soft X-ray counterpart in an otherwise ordinary short GRB shows that long-lasting X-ray emission is likely a common feature of merger-driven bursts and a promising electromagnetic counterpart to gravitational wave sources.

astro-ph.HE↗

arXiv2026

EDCO: Dynamic Curriculum Orchestration for Domain-specific Large Language Model Fine-tuning

Domain-specific large language models (LLMs), typically developed by fine-tuning a pre-trained general-purpose LLM on specialized datasets, represent a significant advancement in applied AI. A common strategy in LLM fine-tuning is curriculum learning, which pre-orders training samples based on metrics like difficulty to improve learning efficiency compared to a random sampling strategy. However, most existing methods for LLM fine-tuning rely on a static curriculum, designed prior to training, which lacks adaptability to the model's evolving needs during fine-tuning. To address this, we propose EDCO, a novel framework based on two key concepts: inference entropy and dynamic curriculum orchestration. Inspired by recent findings that maintaining high answer entropy benefits long-term reasoning gains, EDCO prioritizes samples with high inference entropy in a continuously adapted curriculum. EDCO integrates three core components: an efficient entropy estimator that uses prefix tokens to approximate full-sequence entropy, an entropy-based curriculum generator that selects data points with the highest inference entropy, and an LLM trainer that optimizes the model on the selected curriculum. Comprehensive experiments in communication, medicine and law domains, EDCO outperforms traditional curriculum strategies for fine-tuning Qwen3-4B and Llama3.2-3B models under supervised and reinforcement learning settings. Furthermore, the proposed efficient entropy estimation reduces computational time by 83.5% while maintaining high accuracy.

cs.LG↗

arXiv2025

Video Detective: Seek Critical Clues Recurrently to Answer Question from Long Videos

Long Video Question-Answering (LVQA) presents a significant challenge for Multi-modal Large Language Models (MLLMs) due to immense context and overloaded information, which could also lead to prohibitive memory consumption. While existing methods attempt to address these issues by reducing visual tokens or extending model's context length, they may miss useful information or take considerable computation. In fact, when answering given questions, only a small amount of crucial information is required. Therefore, we propose an efficient question-aware memory mechanism, enabling MLLMs to recurrently seek these critical clues. Our approach, named VideoDetective, simplifies this task by iteratively processing video sub-segments. For each sub-segment, a question-aware compression strategy is employed by introducing a few special memory tokens to achieve purposefully compression. This allows models to effectively seek critical clues while reducing visual tokens. Then, due to history context could have a significant impact, we recurrently aggregate and store these memory tokens to update history context, which would be reused for subsequent sub-segments. Furthermore, to more effectively measure model's long video understanding ability, we introduce GLVC (Grounding Long Video Clues), a long video question-answering dataset, which features grounding critical and concrete clues scattered throughout entire videos. Experimental results demonstrate our method enables MLLMs with limited context length of 32K to efficiently process 100K tokens (3600 frames, an hour-long video sampled at 1fps), requiring only 2 minutes and 37GB GPU memory usage. Evaluation results across multiple long video benchmarks illustrate our method can more effectively seek critical clues from massive information.

cs.CV↗

arXiv2025

Einstein Probe Discovery of an X-ray Flare from K-type Star PM J23221-0301

Stellar flares are an intense stellar activity that can significantly impact the atmospheric composition of the surrounding planets and even the possible existence of life. During such events, the radiative energy of the star is primarily concentrated in the optical and X-ray bands, with the X-ray flux potentially increasing by tens or even hundreds of times. Einstein Probe (EP) detected a new X-ray transient EP J2322.1-0301 on 27 September 2024. Its spatial localization shows a high positional coincidence with the nearby high proper motion K-type star PM J23221-0301. Follow-up X-ray observations confirmed the flux enhancement of the source, while optical spectroscopic monitoring revealed time-variable features, particularly the disappearance of the H-alpha emission line. This X-ray flare is consistent with a characteristic fast-rise-exponential-decay (FRED) light curve, with a rise timescale of 1.4 ks, a decay timescale of 5.7 ks, and a total duration of about 7.1 ks. The peak luminosity in the 0.5-4.0 keV energy band reached about 1.3 x 10^31 erg s^-1, with a total energy release of about 9.1 x 10^34 erg, consistent with the empirical energy correlations observed in magnetic-reconnection-driven stellar flares, as inferred from the multitemperature plasma structure and H-alpha-X-ray energy correlation. This discovery underscores EP's capability in understanding stellar magnetic activity via observing stellar transients.

astro-ph.HE↗

arXiv2025

Investigating Data Pruning for Pretraining Biological Foundation Models at Scale

Biological foundation models (BioFMs), pretrained on large-scale biological sequences, have recently shown strong potential in providing meaningful representations for diverse downstream bioinformatics tasks. However, such models often rely on millions to billions of training sequences and billions of parameters, resulting in prohibitive computational costs and significant barriers to reproducibility and accessibility, particularly for academic labs. To address these challenges, we investigate the feasibility of data pruning for BioFM pretraining and propose a post-hoc influence-guided data pruning framework tailored to biological domains. Our approach introduces a subset-based self-influence formulation that enables efficient estimation of sample importance at low computational cost, and builds upon it two simple yet effective selection strategies, namely Top-k Influence (Top I) and Coverage-Centric Influence (CCI). We empirically validate our method on two representative BioFMs, RNA-FM and ESM-C. For RNA, our framework consistently outperforms random selection baselines under an extreme pruning rate of over 99 percent, demonstrating its effectiveness. Furthermore, we show the generalizability of our framework on protein-related tasks using ESM-C. In particular, our coreset even outperforms random subsets that are ten times larger in both RNA and protein settings, revealing substantial redundancy in biological sequence datasets. These findings underscore the potential of influence-guided data pruning to substantially reduce the computational cost of BioFM pretraining, paving the way for more efficient, accessible, and sustainable biological AI research.

cs.LG↗

arXiv2025

Two-stage primary acceleration in filament initial eruption under a fan-spine magnetic configuration

Understaning the filament rising process is crucial for unveiling the triggering mechanisms of the coronal mass ejections and forecasting the space weather. In this paper, we present a detailed study on the filament initial eruption under a fan-spine structure. It was found that the filament underwent two distinct acceleration stages corresponding to a calss M1.0 and M4.6 flare event, respectively. The first acceleration stage commenced with the filament splitting, after which the upper portion was subsequently heated being a hot channel and slow rose at an average speed of 22 km/s. A set of hot reverse C-shaped loops appeared repeatedly during the filament splitting and a hook structure was recognized at this phase, suggesting ongoing growth of the magnetic flux rope (MFR). When it reached a certain altitude, the hot channel appeared to get into a quasi-static phase with its upper edge seriously decelerated and lower edge expanding downward. Approximately 30 minutes later, as a distinct annular ribbon appeared outside the hook structure, the hot channel rose again at a velocity over 50 km/s accompanied with rapid footpoints drifting, and experienced the second acceleration stage with its axial flux increased to 1.1 X 10^{21} Mx. It is deduced that the filament initial eruption under a magnetic dome possess multi kinetic process. We suggest that the magnetic reconnection taken place within and beneath the filament continues to trigger the growth of pre-eruptive MFR and the first acceleration, when the magnetic reconnection above the filament plays a key role in the second acceleration.

astro-ph.SR↗

arXiv2025

Dexterous Intramyocardial Needle Ablation (d-INA): Design, Fabrication, and In-Vivo Validation

Radiofrequency ablation is widely used to prevent ventricular tachycardia (VT) by creating lesions to inhibit arrhythmias; however, the current surface ablation catheters are limited in creating lesions that are deeper within the left ventricle (LV) wall. Intramyocardial needle ablation (INA) addresses this limitation by penetrating the myocardium and delivering energy from within. Yet, existing INA catheters lack adequate dexterity to navigate the highly asymmetric, trabeculated LV chamber and steer around papillary structures, limiting precise targeting. This work presents a novel dexterous INA (d-INA) toolset designed to enable effective manipulation and creation of deep ablation lesions. The system consists of an outer sheath and an inner catheter, both bidirectionally steerable, along with an integrated ablation needle assembly. Benchtop tests demonstrated that the sheath and catheter reached maximum bending curvatures of 0.088~mm$^{-1}$ and 0.114~mm$^{-1}$, respectively, and achieved stable C-, S-, and non-planar S-shaped configurations. Ex-vivo studies validated the system's stiffness modulation and lesion-creation capabilities. In-vivo experiments in two swine demonstrated the device's ability to reach previously challenging regions such as the LV summit, and achieved a 219\% increase in ablation depth compared with a standard ablation catheter. These results establish the proposed d-INA as a promising platform for achieving deep ablation with enhanced dexterity, advancing VT treatment.

physics.med-ph↗

arXiv2025

MolGA: Molecular Graph Adaptation with Pre-trained 2D Graph Encoder

Molecular graph representation learning is widely used in chemical and biomedical research. While pre-trained 2D graph encoders have demonstrated strong performance, they overlook the rich molecular domain knowledge associated with submolecular instances (atoms and bonds). While molecular pre-training approaches incorporate such knowledge into their pre-training objectives, they typically employ designs tailored to a specific type of knowledge, lacking the flexibility to integrate diverse knowledge present in molecules. Hence, reusing widely available and well-validated pre-trained 2D encoders, while incorporating molecular domain knowledge during downstream adaptation, offers a more practical alternative. In this work, we propose MolGA, which adapts pre-trained 2D graph encoders to downstream molecular applications by flexibly incorporating diverse molecular domain knowledge. First, we propose a molecular alignment strategy that bridge the gap between pre-trained topological representations with domain-knowledge representations. Second, we introduce a conditional adaptation mechanism that generates instance-specific tokens to enable fine-grained integration of molecular domain knowledge for downstream tasks. Finally, we conduct extensive experiments on eleven public datasets, demonstrating the effectiveness of MolGA.

cs.LG↗