arXiv subjects

Chen Shi

Publications and source records attributed to Chen Shi.

At least 37 records · Page 2Linked to original sources

arXiv2026

Conversion Layer Controls the Evolution of Magnetic Deflections Near the Alfven Surface

We examine the statistics of Alfvenic deflections in both sub-Alfvenic and super-Alfvenic solar wind with particular focus on a common parameter that underlies the definition of switchbacks: the magnetic deflection angle. Our findings are in general agreement with earlier studies that suggest magnetic deflection angles > 90 degrees are very unlikely to occur in sub-Alfvenic regimes. We find that their upper limit exhibits an identifiable trend with the Alfven Mach number Ma, suggesting that gradual steepening of Alfvenic deflections with increasing Ma is a plausible mechanism controlling deflection angles in the young solar wind. Further analysis reveals that large velocity fluctuations tend to be important in the largest sub-Alfvenic magnetic deflections with increasing contributions from the parallel component very close to Ma = 1, while virtually no magnetic deflections in the super-Alfvenic regime exhibit such large velocity perturbations. We also determine the local ratio of radial Poynting flux SR to kinetic energy flux KR and find that large sub-Alfvenic deflection angles tend to be dominated by SR, while super-Alfvenic deflections are eventually dominated by the KR associated with the radial solar wind flow. Our results show that within the vicinity of the Alfven surface (where Ma = 1), there is a critical region of parameter space within which velocity deflections approach the Alfven velocity and KR/SR is close to unity. We refer to this region (where | log10(Ma)| < 0.2) as the conversion layer. The conversion layer may play a significant role in the evolution of magnetic defections by providing the medium for converting magnetic energy to particle energy and likely driving the formation of magnetic switchbacks in super-Alfvenic solar wind.

astro-ph.SR↗

arXiv2025

In situ Evidence of 5-minute Oscillations from Parker Solar Probe

The Sun's surface vibrates in characteristic 5-minute oscillations, known as p-modes, generated by sound waves trapped within the convection zone. Although these oscillations have long been hypothesized to reach into the solar wind, direct in situ evidence has remained elusive, even during previous close encounters by Parker Solar Probe (PSP). Here, we present the first promising in situ detection of 5-minute oscillations in the upper solar corona, based on observations from PSP's three closest perihelia. In two events at 9.9 solar radii, we identify statistically significant ($\sim$ 6 $σ$) 3.1-3.2 mHz peaks in the magnetic field power spectrum, each appearing as a large-amplitude, spherically polarized Alfvénic wave train lasting approximately 35 minutes. These results demonstrate that global solar oscillations can reach and potentially influence the solar wind.

astro-ph.SR↗

arXiv2025

UniSplat: Unified Spatio-Temporal Fusion via 3D Latent Scaffolds for Dynamic Driving Scene Reconstruction

Feed-forward 3D reconstruction for autonomous driving has advanced rapidly, yet existing methods struggle with the joint challenges of sparse, non-overlapping camera views and complex scene dynamics. We present UniSplat, a general feed-forward framework that learns robust dynamic scene reconstruction through unified latent spatio-temporal fusion. UniSplat constructs a 3D latent scaffold, a structured representation that captures geometric and semantic scene context by leveraging pretrained foundation models. To effectively integrate information across spatial views and temporal frames, we introduce an efficient fusion mechanism that operates directly within the 3D scaffold, enabling consistent spatio-temporal alignment. To ensure complete and detailed reconstructions, we design a dual-branch decoder that generates dynamic-aware Gaussians from the fused scaffold by combining point-anchored refinement with voxel-based generation, and maintain a persistent memory of static Gaussians to enable streaming scene completion beyond current camera coverage. Extensive experiments on real-world datasets demonstrate that UniSplat achieves state-of-the-art performance in novel view synthesis, while providing robust and high-quality renderings even for viewpoints outside the original camera coverage.

cs.CV↗

arXiv2025

On the Propagation and Damping of Alfvenic Fluctuations in the Outer Solar Corona and Solar Wind

We analyze \textit{Parker Solar Probe} and \textit{Solar Orbiter} observations to investigate the propagation and dissipation of Alfvénic fluctuations from the outer corona to 1~AU. Conservation of wave-action flux provides the theoretical baseline for how fluctuation amplitudes scale with the Alfvén Mach number $M_a$, once solar-wind acceleration is accounted for. Departures from this scaling quantify the net balance between energy injection and dissipation. Fluctuation amplitudes follow wave-action conservation for $M_a < M_a^{b}$ but steepen beyond this break point, which typically lies near the Alfvén surface ($M_a \approx 1$) yet varies systematically with normalized cross helicity $σ_c$ and fluctuation scale. In slow, quasi-balanced streams, the transition occurs at $M_a \lesssim 1$; in fast, imbalanced wind, WKB-like scaling persists to $M_a \gtrsim 1$. Outer-scale fluctuations maintain wave-action conservation to larger $M_a$ than inertial-range modes. The turbulent heating rate $Q$ is largest below $M_a^{b}$, indicating a preferential heating zone shaped by the degree of imbalance. Despite this, the Alfvénic energy flux $F_a$ remains elevated, and the corresponding damping length $Λ_d = F_a/Q$ remains sufficiently large to permit long-range propagation before appreciable damping occurs. Normalized damping lengths $Λ_d/H_A$, where $H_A$ is the inverse Alfvén-speed scale height, are near unity for $M_a \lesssim M_a^{b}$ but decline with increasing $M_a$ and decreasing $U$, implying that incompressible reflection-driven turbulence alone cannot account for the observed dissipation. Additional damping mechanisms -- such as compressible effects -- are likely required to account for the observed heating rates across much of the parameter space.

astro-ph.SR↗

arXiv2025

Two Types of $1/f$ Range in Solar Wind Turbulence

The $1/f$ noise is a ubiquitous phenomenon in natural systems. Since the advent of space exploration, the $1/f$ range has been consistently observed in \textit{in situ} solar wind measurements throughout the heliosphere, sparking decades of debate regarding its origin. Recent Parker Solar Probe (PSP) observations near the Alfvén surface have revealed a systematic absence of the $1/f$ range in pristine solar wind, providing a unique opportunity to investigate its origin in solar wind turbulence. Despite numerous observations of the $1/f$ range at varying frequencies, no study has systematically examined its properties across different solar wind conditions. Here, we identify two distinct types of $1/f$ ranges in solar wind turbulence: the fast/Alfvénic wind type and the slow/mixed wind type. The fast/Alfvénic type appears to be an intrinsic feature of Alfvénic turbulence, while the slow/mixed type resembles classical flicker noise. For the fast/Alfvénic type, we find a near-perfect WKB evolution of the frequency-averaged fluctuation amplitude and an intriguing migration pattern in frequency space. For the slow/mixed type, we examine the solar cycle dependence of the $1/f$ noise using the OMNI-LRO dataset spanning solar cycles 22 to 25. We also analyze the autocorrelation function of the magnetic field vectors and identify a clear relationship between the $1/f$ range and the decline in correlation, as well as unexpected resonance peaks in the autocorrelation function.

astro-ph.SR↗

arXiv2025

CX-Mind: A Pioneering Multimodal Large Language Model for Interleaved Reasoning in Chest X-ray via Curriculum-Guided Reinforcement Learning

Chest X-ray (CXR) imaging is one of the most widely used diagnostic modalities in clinical practice, encompassing a broad spectrum of diagnostic tasks. Recent advancements have seen the extensive application of reasoning-based multimodal large language models (MLLMs) in medical imaging to enhance diagnostic efficiency and interpretability. However, existing multimodal models predominantly rely on "one-time" diagnostic approaches, lacking verifiable supervision of the reasoning process. This leads to challenges in multi-task CXR diagnosis, including lengthy reasoning, sparse rewards, and frequent hallucinations. To address these issues, we propose CX-Mind, the first generative model to achieve interleaved "think-answer" reasoning for CXR tasks, driven by curriculum-based reinforcement learning and verifiable process rewards (CuRL-VPR). Specifically, we constructed an instruction-tuning dataset, CX-Set, comprising 708,473 images and 2,619,148 samples, and generated 42,828 high-quality interleaved reasoning data points supervised by clinical reports. Optimization was conducted in two stages under the Group Relative Policy Optimization framework: initially stabilizing basic reasoning with closed-domain tasks, followed by transfer to open-domain diagnostics, incorporating rule-based conditional process rewards to bypass the need for pretrained reward models. Extensive experimental results demonstrate that CX-Mind significantly outperforms existing medical and general-domain MLLMs in visual understanding, text generation, and spatiotemporal alignment, achieving an average performance improvement of 25.1% over comparable CXR-specific models. On real-world clinical dataset (Rui-CXR), CX-Mind achieves a mean recall@1 across 14 diseases that substantially surpasses the second-best results, with multi-center expert evaluations further confirming its clinical utility across multiple dimensions.

cs.LG↗

arXiv2025

Action Unit Enhance Dynamic Facial Expression Recognition

Dynamic Facial Expression Recognition(DFER) is a rapidly evolving field of research that focuses on the recognition of time-series facial expressions. While previous research on DFER has concentrated on feature learning from a deep learning perspective, we put forward an AU-enhanced Dynamic Facial Expression Recognition architecture, namely AU-DFER, that incorporates AU-expression knowledge to enhance the effectiveness of deep learning modeling. In particular, the contribution of the Action Units(AUs) to different expressions is quantified, and a weight matrix is designed to incorporate a priori knowledge. Subsequently, the knowledge is integrated with the learning outcomes of a conventional deep learning network through the introduction of AU loss. The design is incorporated into the existing optimal model for dynamic expression recognition for the purpose of validation. Experiments are conducted on three recent mainstream open-source approaches to DFER on the principal datasets in this field. The results demonstrate that the proposed architecture outperforms the state-of-the-art(SOTA) methods without the need for additional arithmetic and generally produces improved results. Furthermore, we investigate the potential of AU loss function redesign to address data label imbalance issues in established dynamic expression datasets. To the best of our knowledge, this is the first attempt to integrate quantified AU-expression knowledge into various DFER models. We also devise strategies to tackle label imbalance, or minor class problems. Our findings suggest that employing a diverse strategy of loss function design can enhance the effectiveness of DFER. This underscores the criticality of addressing data imbalance challenges in mainstream datasets within this domain. The source code is available at https://github.com/Cross-Innovation-Lab/AU-DFER.

cs.CV↗

arXiv2025

Robust Estimation of Double Autoregressive Models via Normal Mixture QMLE

This paper investigates the estimation of the double autoregressive (DAR) model in the presence of skewed and heavy-tailed innovations. We propose a novel Normal Mixture Quasi-Maximum Likelihood Estimation (NM-QMLE) method to address the limitations of conventional quasi-maximum likelihood estimation (QMLE) under non-Gaussian conditions. By incorporating a normal mixture distribution into the quasi-likelihood framework, NM-QMLE effectively captures both heavy-tailed behavior and skewness. A critical contribution of this paper is addressing the often-overlooked challenge of selecting the appropriate number of mixture components, $K$, a key parameter that significantly impacts model performance. We systematically evaluate the effectiveness of different model selection criteria. Under regularity conditions, we establish the consistency and asymptotic normality of the NM-QMLE estimator for DAR($p$) models. Numerical simulations demonstrate that NM-QMLE outperforms commonly adopted QMLE methods in terms of estimation accuracy, particularly when the innovation distribution deviates from normality. Our results also show that while criteria like BIC and ICL improve parameter estimation of $K$, fixing a small order of components provides comparable accuracy. To further validate its practical applicability, we apply NM-QMLE to empirical data from the S\&P 500 index and assess its performance through Value at Risk (VaR) estimation. The empirical findings highlight the effectiveness of NM-QMLE in modeling real-world financial data and improving risk assessment. By providing a robust and flexible estimation approach, NM-QMLE enhances the analysis of time series models with complex innovation structures, making it a valuable tool in econometrics and financial modeling.

stat.ME↗

arXiv2025

DriveX: Omni Scene Modeling for Learning Generalizable World Knowledge in Autonomous Driving

Data-driven learning has advanced autonomous driving, yet task-specific models struggle with out-of-distribution scenarios due to their narrow optimization objectives and reliance on costly annotated data. We present DriveX, a self-supervised world model that learns generalizable scene dynamics and holistic representations (geometric, semantic, and motion) from large-scale driving videos. DriveX introduces Omni Scene Modeling (OSM), a module that unifies multimodal supervision-3D point cloud forecasting, 2D semantic representation, and image generation-to capture comprehensive scene evolution. To simplify learning complex dynamics, we propose a decoupled latent world modeling strategy that separates world representation learning from future state decoding, augmented by dynamic-aware ray sampling to enhance motion modeling. For downstream adaptation, we design Future Spatial Attention (FSA), a unified paradigm that dynamically aggregates spatiotemporal features from DriveX's predictions to enhance task-specific inference. Extensive experiments demonstrate DriveX's effectiveness: it achieves significant improvements in 3D future point cloud prediction over prior work, while attaining state-of-the-art results on diverse tasks including occupancy prediction, flow estimation, and end-to-end driving. These results validate DriveX's capability as a general-purpose world model, paving the way for robust and unified autonomous driving frameworks.

cs.CV↗

arXiv2025

C-learning in estimation of optimal individualized treatment regimes for recurrent disease

Recurrent events, characterized by the repeated occurrence of the same event in an individual, are a common type of data in medical research. Motivated by cancer recurrences, we aim to estimate the optimal individualized treatment regime (ITR) that effectively mitigates such recurrent events. An ITR is a decision rule that assigns the optimal treatment to each patient, based on personalized information, with the aim of maximizing the overall therapeutic benefits. However, existing studies of estimating ITR mainly focus on first-time events rather than recurrent events. To address the issue of determining the optimal ITR for recurrent events, we propose the Recurrent C-learning (ReCL) method to identify the optimal ITR from two or multiple treatment options. The proposed method reformulates the optimization problem into a weighted classification problem. We introduce three estimators for the misclassification cost: the outcome regression estimator, the inverse probability weighting estimator, and the augmented inverse probability weighting estimator. The ReCL method leverages classification techniques to generate an interpretable optimal ITR tailored for recurrent event data. The advantages of the ReCL method are demonstrated through simulations under various scenarios. Furthermore, based on real data on colorectal cancer treatments, we employ this novel method to derive interpretable tree treatment regimes for colorectal cancer, thus providing a practical framework for enhancing treatment strategies.

stat.ME↗

arXiv2025

Evolution of MHD turbulence in the expanding solar wind: residual energy and intermittency

We conduct 3D magnetohydrodynamic (MHD) simulations of decaying turbulence in the solar wind context. To account for the spherical expansion of the solar wind, we implement the expanding box model. The initial turbulence comprises uncorrelated counter-propagating Alfvén waves and exhibits an isotropic power spectrum. Our findings reveal the consistent generation of negative residual energy whenever nonlinear interactions are present, independent of the normalized cross helicity $σ_c$ and compressibility. The spherical expansion facilitates this process. The resulting residual energy is primarily distributed in the perpendicular direction, with $[S_2(\mathbf{b})-S_2(\mathbf{u})] \propto l_\perp$ or equivalently $-E_r \propto k_\perp^{-2}$. Here $S_2(\mathbf{b})$ and $S_2(\mathbf{u})$ are second-order structure functions of magnetic field and velocity respectively. In most runs, $S_2(\mathbf{b})$ develops a scaling relation $S_2(\mathbf{b}) \propto l_\perp^{1/2}$ ($E_b \propto k_\perp^{-3/2}$). In contrast, $S_2(\mathbf{u})$ is consistently shallower than $S_2(\mathbf{b})$, which aligns with in-situ observations of the solar wind. We observe that the higher-order statistics of the turbulence, which act as a proxy for intermittency, depend on the initial $σ_c$ and are strongly affected by the expansion effect. Generally, the intermittency is more pronounced when the expansion effect is present. Finally, we find that in our simulations although the negative residual energy and intermittency grow simultaneously as the turbulence evolves, the causal relation between them seems to be weak, possibly because they are generated on different scales.

astro-ph.SR↗

arXiv2024

Dominance of 2-Minute Oscillations near the Alfvén Surface

Alfvén waves, considered one of the primary candidates for heating and accelerating the fast solar wind, are ubiquitous in spacecraft observations, yet their origin remains elusive. In this study, we analyze data from the first 19 encounters of the Parker Solar Probe (PSP) and report dominance of 2-minute oscillations near the Alfvén surface. The frequency-rectified trace magnetic power spectral density (PSD) of these oscillations indicates that the fluctuation energy is concentrated around 2 minutes for the ``youngest'' solar wind. Further analysis using wavelet spectrograms reveals that these oscillations primarily consist of outward-propagating, spherically polarized Alfvén wave bursts. Through Doppler analysis, we show that the wave frequency observed in the spacecraft frame can be mapped directly to the launch frequency at the base of the corona, where previous studies have identified a distinct peak around 2 minutes ($\sim 8$ mHz) in the spectrum of swaying motions of coronal structures observed by SDO AIA. These findings strongly suggest that the Alfvén waves originate from the solar atmosphere. Furthermore, statistical analysis of the PSD deformation beyond the Alfvén surface supports the idea of dynamic formation of the otherwise absent $1/f$ range in the solar wind turbulence spectrum.

astro-ph.SR↗

arXiv2024

In situ observations of large amplitude Alfvén waves heating and accelerating the solar wind

After leaving the Sun's corona, the solar wind continues to accelerate and cools, but more slowly than expected for a freely expanding adiabatic gas. We use in situ measurements from the Parker Solar Probe and Solar Orbiter spacecrafts to investigate a stream of solar wind as it traverses the inner heliosphere. The observations show heating and acceleration of the the plasma between the outer edge of the corona and near the orbit of Venus, in connection to the presence of large amplitude Alfvén waves. Alfvén waves are perturbations in the interplanetary magnetic field that transport energy. Our calculations show the damping and mechanical work performed by the Alfvén waves is sufficient to power the heating and acceleration of the fast solar wind in the inner heliosphere.

astro-ph.SR↗

arXiv2024

Solar Wind Structures from the Gaussianity of Magnetic Magnitude

One of the primary science objectives of Parker Solar Probe (PSP) is to determine the structures and dynamics of the plasma and magnetic fields at the sources of the solar wind. However, establishing the connection between {\it in situ} measurements and structures and dynamics in the solar atmosphere is challenging: most of the magnetic footpoint mapping techniques have significant uncertainties in the source localization of a plasma parcel observed {\it in situ}, and the PSP plasma measurements suffer from a limited field of view. Therefore it is of interest to investigate whether {\it in situ} measurements can be used on their own to identify streams originating from the same structures in the corona more finely than the well known fast wind-coronal hole, slow wind-elsewhere distinction. Here we develop a novel time series visualization method \textcolor{red}{(time-frequency representation or TFR)} named Gaussianity Scalogram. Utilizing this method, by analyzing the magnetic magnitude data from both PSP and Ulysses, we successfully identify {\it in situ} structures that are possible remnants of solar atmospheric and magnetic structures spanning more than seven orders of magnitude, from years to seconds, including polar and mid-latitude coronal holes, as well as structures compatible with super-granulation , ``jetlets'' and ``picoflares''. \textcolor{red}{Furthermore, computer simulations of Alfvénic turbulence successfully reproduce the Gaussianization of the magnetic magnitude for locally homogeneous structures.} Building upon these discoveries, the Gaussianity Scalogram can help future studies to reveal the fractal-like fine structures in the solar wind time series from both PSP and decades-old data archive.

physics.space-ph↗

arXiv2024

Scale-Dependent Dynamic Alignment in MHD Turbulence: Insights into Intermittency, Compressibility, and Imbalance Effects

Scale-Dependent Dynamic Alignment (SDDA) in Elsässer field fluctuations is theorized to suppress nonlinearities and modulate the energy spectrum. Limited empirical evidence exists for SDDA within the solar wind turbulence's inertial range. We analyzed data from the WIND mission to assess the effects of compressibility, intermittency, and imbalance on SDDA. SDDA consistently appears at energy-containing scales, with a trend toward misalignment at inertial scales. Compressible fluctuations show no increased alignment; however, their impact on SDDA's overall behavior is minimal. The alignment angles inversely correlate with field gradient intensity, likely due to "anomalous" or "counterpropagating" wave packet interactions. This suggests that SDDA originates from mutual shearing of Elsässer fields during imbalanced ($δ\boldsymbol{z}^{\pm} \gg δ\boldsymbol{z}^{\mp}$) interactions. Rigorous thresholding on field gradient intensity reveals SDDA signatures across much of the inertial range. The scaling of Elsässer increments' alignment angle, $Θ^{z}$, steepens with increasing global Alfvénic imbalance, while the angle between magnetic and velocity field increments, $Θ^{ub}$, becomes shallower. $Θ^{ub}$ only correlates with global Elsässer imbalance, steepening as the imbalance increases. Furthermore, increasing alignment in $Θ^{ub}$ persists deep into the inertial range of balanced intervals but collapses at large scales for imbalanced ones. Simplified theoretical analysis and modeling of high-frequency, low-amplitude noise in the velocity field indicate significant impacts on alignment angle measurements even at very low frequencies, with effects growing as global imbalance increases.

physics.space-ph↗

arXiv2024

Higher-Order Analysis of Three-Dimensional Anisotropy in Imbalanced Alfvénic Turbulence

We analyze in-situ observations of imbalanced solar wind turbulence to evaluate MHD turbulence models grounded in "Critical Balance" (CB) and "Scale-Dependent Dynamic Alignment" (SDDA). At energy injection scales, both outgoing and ingoing modes exhibit a weak cascade; a simultaneous tightening of SDDA is noted. Outgoing modes persist in a weak cascade across the inertial range, while ingoing modes shift to a strong cascade at $λ\approx 3 \times 10^{4} d_i$, with associated spectral scalings deviating from expected behavior due to "anomalous coherence" effects. The inertial range comprises two distinct sub-inertial segments. Beyond $λ\gtrsim 100 d_i$, eddies adopt a field-aligned tube topology, with SDDA signatures mainly evident in high amplitude fluctuations. The scaling exponents $ζ_{n}$ of the $n$-th order conditional structure functions, orthogonal to both the local mean field and fluctuation direction, align with the analytical models of Chandran et al. 2015 and Mallet et al. 2017, indicating "multifractal" statistics and strong intermittency; however, scaling in parallel and displacement components is more concave than predicted, possibly influenced by expansion effects. Below $λ\approx 100 d_i$, eddies become increasingly anisotropic, evolving into thin current sheet-like structures. Concurrently, $ζ_{n}$ scales linearly with order, marking a shift towards "monofractal" statistics. At $λ\approx 8 d_i$, the increase in aspect ratio halts, and the eddies become quasi-isotropic. This change may signal tearing instability, leading to reconnection, or result from energy redirection into the ion-cyclotron wave spectrum, aligning with the "helicity barrier". Our analysis utilizes 5-point structure functions, proving more effective than the traditional 2-point method in capturing steep scaling behaviors at smaller scales.

physics.space-ph↗

arXiv2024

Analytic Model and Magnetohydrodynamic Simulations of Three-dimensional Magnetic Switchbacks

Parker Solar Probe observations reveal that the near-Sun space is almost filled with magnetic switchbacks (``switchbacks'' hereinafter), which may be a major contributor to the heating and acceleration of solar wind. Here, for the first time, we develop an analytic model of an axisymmetric switchback with uniform magnetic field strength. In this model, three parameters control the geometry of the switchback: height (length along the background magnetic field), width (thickness along radial direction perpendicular to the background field), and the radial distance from the center of switchback to the central axis, which is a proxy of the size of the switchback along the third dimension. We carry out three-dimensional magnetohydrodynamic simulations to investigate the dynamic evolution of the switchback. Comparing simulations conducted with compressible and incompressible codes, we verify that compressibility, i.e. parametric decay instability, is necessary for destabilizing the switchback. Our simulations also reveal that the geometry of the switchback significantly affects how fast the switchback destabilizes. The most stable switchbacks are 2D-like (planar) structures with large aspect ratios (length to width), consistent with the observations. We show that when plasma beta ($β$) is smaller than one, the switchback is more stable as $β$ increases. However, when $β$ is greater than one, the switchback becomes very unstable as the pattern of the growing compressive fluctuations changes. Our results may explain some of the observational features of switchbacks, including the large aspect ratios and nearly constant occurrence rates in the inner heliosphere.

physics.space-ph↗

arXiv2024

AlphaFin: Benchmarking Financial Analysis with Retrieval-Augmented Stock-Chain Framework

The task of financial analysis primarily encompasses two key areas: stock trend prediction and the corresponding financial question answering. Currently, machine learning and deep learning algorithms (ML&DL) have been widely applied for stock trend predictions, leading to significant progress. However, these methods fail to provide reasons for predictions, lacking interpretability and reasoning processes. Also, they can not integrate textual information such as financial news or reports. Meanwhile, large language models (LLMs) have remarkable textual understanding and generation ability. But due to the scarcity of financial training datasets and limited integration with real-time knowledge, LLMs still suffer from hallucinations and are unable to keep up with the latest information. To tackle these challenges, we first release AlphaFin datasets, combining traditional research datasets, real-time financial data, and handwritten chain-of-thought (CoT) data. It has a positive impact on training LLMs for completing financial analysis. We then use AlphaFin datasets to benchmark a state-of-the-art method, called Stock-Chain, for effectively tackling the financial analysis task, which integrates retrieval-augmented generation (RAG) techniques. Extensive experiments are conducted to demonstrate the effectiveness of our framework on financial analysis.

cs.CL↗