arXiv subjects

Lan Wang

Publications and source records attributed to Lan Wang.

At least 19 recordsLinked to original sources

arXiv2026

Sparse Additive Off-Policy Evaluation for Reinforcement Learning with Potentially Limited Number of Trajectories

We develop a new framework for flexible, nonlinear, and interpretable off-policy evaluation for infinite-horizon reinforcement learning. To handle large state spaces and support transparent decision-making, we model the Q-function using a nonlinear function class with a sparse additive structure. We derive high-probability finite-sample error bounds for estimating the value function of a target policy and show that the bounds depend only logarithmically on the ambient dimension $d$, thereby alleviating the curse of dimensionality. In contrast to most existing theory for off-policy evaluation, which typically assumes access to many trajectories, our analysis guarantees accurate value estimation when either the number of trajectories or the time horizon is sufficiently large. In addition, we propose a group-sparsity-based feature screening procedure that identifies, with high probability, a reduced feature set containing all relevant covariates. Numerical experiments demonstrate the effectiveness of the proposed approach.

stat.ML↗

arXiv2026

Temporal Leakage in Financial News NLP: A Multi-Architecture Audit with a Regime-Specific M&A Signal

Financial-news direction prediction has become a popular NLP benchmark, yet reported gains depend critically on whether the train-test split is chronological or random, i.e., on temporal leakage. We audit this dependence on a 49,799-article corpus across 16 feature-model combinations spanning TF-IDF, MiniLM, FinBERT, and fine-tuned RoBERTa-large / DeBERTa-v3-large, plus separate zero/few-shot and LoRA probes of Llama-3 and Qwen2.5 LLMs: random splits inflate MCC by $1.1\times$ to $6.5\times$, tracking model capacity and feature richness, and end-to-end FinBERT fine-tuning re-amplifies rather than closes the gap (size-matched ratio $1.75\times$). Conditioning on event type, mergers and acquisitions (M&A) is the only audited category with a positive locked-test signal under near-temporal chronological evaluation (TF-IDF MCC $= 0.138$ train-only, $0.068$ under train$\cup$val refit; 10,000-permutation $p < 10^{-3}$); the signal does not transfer to FNSPID's 2009-2020 U.S. corpus, localising the headline to our 2024-2025 European-tilted M&A semantics rather than a universal predictor. Three independent role labellers converge on acquirer-tagged articles as the signal locus, a power-limited qualitative convergence rather than a hypothesis-tested asymmetry. Chronological splitting plays for financial NLP the role characteristics-purging plays for asset pricing: it strips the predictable, stale component of news and leaves a residual that is small, event-localized, and lexically shallow. We advocate leakage audits as a required disclosure for financial-NLP benchmarks.

cs.CL↗

arXiv2026

How mergers shape galaxy morphology in the IllustrisTNG simulation

How galaxy mergers drive morphological evolution remains an open question. Traditional views hold that major mergers produce elliptical galaxies, while minor mergers form dispersion-dominated components such as galactic bulges. However, more recent work has challenged this simple picture, suggesting a more complex evolutionary scenario. In this study, we use the IllustrisTNG cosmological simulation to investigate how mergers shape galaxy morphology, across a broad range of galaxy masses and merger mass ratios. Our results show that the post-merger galaxy morphology is primarily determined by three factors: collision angle $\overline{\theta}$, cold gas fraction $f_\mathrm{cold\,gas}$, and pre-merger galaxy morphology $\mathrm{(B/T)_{*,pre}}$. Specifically, spiral-in mergers with large $\overline{\theta}$ increase the rotational support of the system, allowing the gas to settle into an extended disk. When the system is rich in cold gas, star formation within the newly formed gas disk can further strengthen the disk-dominated structure of the remnant. On the other hand, head-on mergers with small $\overline{\theta}$ typically disrupt ordered galactic motion, producing more dispersion-supported remnants. Overall, we interpret our results within a unified picture of merger-driven morphological transformation in galaxies.

astro-ph.GA↗

arXiv2026

From Single- to Cross-Document: Benchmarking Multi-Granularity Event Analysis of Large Language Models

Event analysis is an essential and fundamental direction of information extraction, involving various event-centric tasks at different granularity of documents. While large language models (LLMs) have preliminarily achieved promising performance in part of these tasks individually, their capability in event analysis still lacks comprehensive understanding due to restricted document granularity, task designs, and data source of existing benchmarks. To address these limitations, we introduce MiGUE-Bench, a systematic benchmark for assessing the performance of LLMs in multi-granularity event analysis. To support large-scale evaluation, we first develop an LLM-driven self-correcting annotation framework called MiGUE-Pipeline, enabling scalable acquisition of high-quality source data of events with automatic labels. Then, we design four core tasks in our benchmark, i.e., event detection, relation reasoning, structure induction, and future prediction, to probe model competence at different levels, from atomic event details to complex cross-document narratives. Extensive experiments on state-of-the-art LLMs and retrieval-augmented generation (RAG) methods delineate the current capability boundary and identify critical deficiencies, providing insights into the future improvement of LLMs in challenging event analysis tasks.

cs.CL↗

arXiv2026

QSVideo: Query-Conditioned Semantic Temporal Retrieval for Video Understanding

The performance of vision-language models (VLMs) in video understanding declines with increasing video duration, as video moments unrelated to the query confuse their language components. Multimodal retrieval has emerged as a critical component of video understanding, addressing this challenge by localizing key visual evidence. However, existing multimodal retrieval methods suffer from biased relevance estimation, limited diversity, and temporal collapse. In this paper, we propose QSVideo, a unified framework that systematically addresses relevance, diversity, and temporal modeling in video retrieval. We first introduce a query-conditioned semantic ranker, QSRanker, which reformulates arbitrary questions into retrieval-friendly queries and estimates structured relevance along object, action, and location dimensions. Building upon this, we design QSRetrieval to jointly optimize relevance and diversity for more informative frame selection. Moreover, we propose temporal alignment strategies tailored for both long and streaming videos to improve evidence recall. Extensive experiments on long and streaming video benchmarks demonstrate that QSVideo greatly enhances video VLM performance under strict frame limit constraints. The code is available at https://github.com/human-analysis/QSVideo.

cs.CV↗

arXiv2026

Deep neural network yields regularization for ill-posed inverse problems

This paper studies the regularization of ill-posed inverse problems by deep neural networks (DNNs). We extend architecture-based regularization from shallow networks to deep models by developing a deterministic framework in which the admissible network class is enlarged adaptively and the resulting architecture complexity acts as the regularization mechanism. We propose two discrepancy-principle-driven expanding DNN algorithms to treat the cases where an explicit parameter-radius bound is available and unavailable, respectively. For both algorithms, we prove the finite termination of the adaptive expansion procedure and the convergence of the regularized solutions as the noise level vanishes. In addition, we derive explicit asymptotic bounds on the terminal network architecture, thereby quantifying how the required network complexity scales with the noise level. Numerical experiments on several representative linear and non-linear inverse problems support the theoretical findings and illustrate the practical usefulness of the proposed framework.

math.NA↗

arXiv2026

From Map-and-Encap to BIER: Observations on Network Routing Scalability

The TCP/IP protocol stack uses IP addresses for two distinct roles: identifying hosts and locating their attachment points in the network topology. This dual purpose creates a fundamental tension that has led to routing and forwarding scalability challenges throughout the history of the Internet in unicast packet delivery and, more notably, in multicast delivery. This paper reviews the evolution of routing scalability solutions over the years and makes four observations. First, map-and-encap is a recurring architectural solution shared by all scalable unicast and multicast delivery methods, developed independently across different problem contexts. Second, a new solution tends to succeed when it can bring immediate local gains to early adopters without requiring coordination across administrative domains. Third, network routing and forwarding designs that depend on external factors, such as the number of distinct end sites or even application-specific deliveries, inherently preclude an upper bound on their scalability. Fourth, today's inter-domain routing protocol, BGP, lacks a topological abstraction equivalent to an egress router within a routing domain, thereby inherently preventing a map-and-encap solution for scalability. These observations offer insights into the design of future scalable routing system architectures.

cs.NI↗

arXiv2026

From Speech to Profile: A Protocol-Driven LLM Agent for Psychological Profile Generation

The psychological profile that structurally documents the case of a depression patient is essential for psychotherapy. Large language models can be applied to summarize the profiles from counseling speech, however, it may suffer from long-context forgetting and produce unverifiable hallucinations, due to overlong length of speech, multi-party interactions and unstructured chatting. Hereby, we propose a StreamProfile, a streaming framework that processes counseling speech incrementally, extracts evidences grounded from ASR transcriptions by storing it in a Hierarchical Evidence Memory, and then performs a Chain-of-Thought pipeline according to PM+ psychological intervention for clinical reasoning. The final profile is synthesized strictly from those evidences, making every claim traceable. Experiments on real-world teenager counseling speech have shown that the proposed StreamProfile system can accurately generate the profiles and prevent hallucination.

cs.SD↗

arXiv2026

Learning to Attend to Depression-Related Patterns: An Adaptive Cross-Modal Gating Network for Depression Detection

Automatic depression detection using speech signals with acoustic and textual modalities is a promising approach for early diagnosis. Depression-related patterns exhibit sparsity in speech: diagnostically relevant features occur in specific segments rather than being uniformly distributed. However, most existing methods treat all frames equally, assuming depression-related information is uniformly distributed and thus overlooking this sparsity. To address this issue, we proposes a depression detection network based on Adaptive Cross-Modal Gating (ACMG) that adaptively reassigns frame-level weights across both modalities, enabling selective attention to depression-related segments. Experimental results show that the depression detection system with ACMG outperforms baselines without it. Visualization analyses further confirm that ACMG automatically attends to clinically meaningful patterns, including low-energy acoustic segments and textual segments containing negative sentiments.

cs.SD↗

arXiv2026

Why Are Some Optically Red Spirals NUV-r Blue?

To understand the complicated formation processes of disk galaxies, we carry out a comparative study for NUV-r blue and red spiral galaxies drawn from a parent sample of u-r red spirals with $M_{*} > 10^{10.5} M_{\odot}$ at 0.02 < z < 0.07, based on the optical data from the Sloan Digital Sky Survey (SDSS) and the ultraviolet (UV) data from the Galaxy Evolution Explorer (GALEX). The analyses of the images and surface brightness profiles in the NUV and optical bands show that the differences between NUV-r blue and red spirals mainly occur in the outer disks (1-3 $R_{\rm e}$), and the contrast in NUV band is much larger than that in the optical bands. Both the positions on the star formation main sequence diagram and the NUV-r color profiles suggest that NUV-r red spirals have been fully quenched, whereas NUV-r blue spirals host quenched bulges and inner disks, as well as star-forming outer disks. Particularly, the disk mass-size relations indicate that, at a given disk mass, NUV-r blue spirals possess larger optical disks than NUV-r red spirals, by a factor of $\sim 1.20$. The environments and optical morphologies are consistent with the scenario that NUV-r blue spirals obtained fresh fuel for star formation either by interacting or merging with gas-rich galaxies or through accreting surrounding HI gas.

astro-ph.GA↗

arXiv2026

Looping Back to Move Forward: Recursive Transformers for Efficient and Flexible Large Multimodal Models

Large Multimodal Models (LMMs) have achieved remarkable success in vision-language tasks, yet their vast parameter counts are often underutilized during both training and inference. In this work, we embrace the idea of looping back to move forward: reusing model parameters through recursive refinement to extract stronger multimodal representations without increasing model size. We propose RecursiveVLM, a recursive Transformer architecture tailored for LMMs. Two key innovations enable effective looping: (i) a Recursive Connector that aligns features across recursion steps by fusing intermediate-layer hidden states and applying modality-specific projections, respecting the distinct statistical structures of vision and language tokens; (ii) a Monotonic Recursion Loss that supervises every step and guarantees performance improves monotonically with recursion depth. This design transforms recursion into an on-demand refinement mechanism: delivering strong results with few loops on resource-constrained devices and progressively improving outputs when more computation resources are available. Experiments show consistent gains of +3% over standard Transformers and +7% over vanilla recursive baselines, demonstrating that strategic looping is a powerful path toward efficient, deployment-adaptive LMMs.

cs.LG↗

arXiv2026

Double Fairness Policy Learning: Integrating Action Fairness and Outcome Fairness in Decision-making

Fairness is a central pillar of trustworthy machine learning, especially in domains where accuracy- or profit-driven optimization is insufficient. While most fairness research focuses on supervised learning, fairness in policy learning remains less explored. Because policy learning is interventional, it induces two distinct fairness targets: action fairness (equitable action assignments) and outcome fairness (equitable downstream consequences). Crucially, equalizing actions does not generally equalize outcomes when groups face different constraints or respond differently to the same action. We propose a novel double fairness learning (DFL) framework that explicitly manages the trade-off among three objectives: action fairness, outcome fairness, and value maximization. We integrate fairness directly into a multi-objective optimization problem for policy learning and employ a lexicographic weighted Tchebyshev method that recovers Pareto solutions beyond convex settings, with theoretical guarantees on the regret bounds. Our framework is flexible and accommodates various commonly used fairness notions. Extensive simulations demonstrate improved performance relative to competing methods. In applications to a motor third-party liability insurance dataset and an entrepreneurship training dataset, DFL substantially improves both action and outcome fairness while incurring only a modest reduction in overall value.

stat.ML↗

arXiv2025

SRM at 30: Lessons from Early Data-Centric Networking and Their Impact on Named Data Networking

A 1995 SIGCOMM paper, "A Reliable Multicast Framework for Light-weight Sessions and Application-Level Framing", commonly known as SRM, explored a fundamentally new approach to reliable multiparty data delivery. Rather than adapting established sender-driven reliable unicast mechanisms to multicast, as most contemporaneous proposals did, SRM introduced a data-centric model in which data receivers recover losses by explicitly requesting missing data. Thirty years later, we revisit the SRM framework, examining the challenges it faced, the lessons learned, and its influence on the later development of Named Data Networking (NDN). Experimentations with SRM revealed a fundamental semantic mismatch between its data-centric framework and IP's address-based delivery; while the application layer named data, the network layer remained 'blind' to those names, resulting in inefficient loss recovery. NDN resolves this architectural friction by aligning network delivery with the data-retrieval model and by securing data directly rather than securing communication channels. This retrospective highlights how early insights from SRM informed key design decisions in NDN and illustrates how NDN's design emerged from the cumulative insights gained over decades of networking research and development.

cs.NI↗

arXiv2025

SeedProteo: Accurate De Novo All-Atom Design of Protein Binders

We present SeedProteo, a diffusion-based model for de novo all-atom protein design. We demonstrate how to repurpose a cutting-edge folding architecture into a powerful generative design framework by effectively integrating self-conditioning features. Extensive benchmarks highlight the model's capabilities across two distinct tasks: in unconditional generation, SeedProteo exhibits superior length generalization and structural diversity, maintaining robustness for long sequences and complex topologies; in binder design, it achieves state-of-the-art performance among open-source methods, attaining the highest in-silico design success rates, structural diversity and novelty. Finally, we validate SeedProteo through wet-lab assays on two therapeutic targets, achieving hit rates of 70%-80% and picomolar-level binding affinities, establishing leading results. To facilitate community adoption, we provide public access to SeedProteo via a webserver (https://seedfold.io/proteinDesign).

q-bio.BM↗

arXiv2025

SeedFold: Scaling Biomolecular Structure Prediction

Highly accurate biomolecular structure prediction is a key component of developing biomolecular foundation models, and one of the most critical aspects of building foundation models is identifying the recipes for scaling the model. In this work, we present SeedFold, a folding model that successfully scales up the model capacity. Our contributions are threefold: first, we identify an effective width-scaling strategy for the Pairformer to increase representation capacity; second, we introduce a novel linear triangular attention that reduces computational complexity to enable efficient scaling; finally, we construct a large-scale distillation dataset to substantially enlarge the training set. Experiments on FoldBench show that SeedFold outperforms AlphaFold3 on most protein-related tasks.

q-bio.BM↗

arXiv2025

X-Talk: On the Underestimated Potential of Modular Speech-to-Speech Dialogue System

We present X-Talk, an open-source framework that champions a decoupled, modular design for LLM-driven speech-to-speech (S2S) systems. While the dominant trend favors end-to-end (E2E) modeling to optimize information flow, these "omni-models" often struggle to balance the competing objectives of complex speech tasks within a single network. X-Talk challenges this paradigm by demonstrating that a systematically optimized cascaded pipeline can achieve sub-second latency without sacrificing modular flexibility. Our framework seamlessly integrates specialized front-end components (e.g., VAD, speech enhancement) and diverse understanding models (e.g., ASR, emotion, and environmental sound analysis) with LLM capabilities like retrieval-augmented generation (RAG) and tool use. By revitalizing the cascaded approach, X-Talk highlights the underestimated potential of modular S2S systems and provides a robust foundation for future research and applications.

cs.SD↗

arXiv2025

OrdMoE: Preference Alignment via Hierarchical Expert Group Ranking in Multimodal Mixture-of-Experts LLMs

Preference learning has recently emerged as a pivotal strategy for post-training alignment of Multimodal Large Language Models (MLLMs). However, existing approaches predominantly rely on external human-annotated preference data, which is costly and labor-intensive to collect. In this work, we propose OrdMoE, a novel preference alignment framework that bypasses the reliance on external human preferences entirely by leveraging intrinsic signals within Mixture-of-Experts (MoE) architectures. Specifically, we observe that the router's expert selection scores implicitly encode a quality-aware ranking of responses (i.e. higher-scoring experts consistently generate higher-quality outputs). Building on this insight, OrdMoE constructs an internal preference hierarchy by grouping experts into ranked tiers based on their per-token routing scores and activating each tier separately to produce a sequence of responses with increasing quality. This yields a zero-cost, self-supervised preference ordering over generated responses, which can be directly optimized using standard preference learning objectives. Extensive experiments across multiple multimodal benchmarks demnstrate that OrdMoE significantly enhances both alignment and overall performance of multimodal Mixture-of-Experts LLMs, achieving competitive results without requiring any human-annotated preference data.

cs.LG↗

arXiv2025

Comparing galaxy merger orbits in hydrodynamical simulation and in dark-matter-only simulation

To investigate how the presence of baryons in simulations affects galaxy merger orbits, we compare in detail the merger timescales and orbits of the matched merger pairs in TNG100 hydrodynamical simulations and their corresponding dark-matter-only simulations, for different resolution levels. Compared with the mergers in the TNG100-1-Dark simulation without baryons, the matched mergers in the TNG100-1 simulation have similar infall time, but have statistically earlier merger times and therefore shorter merger timescales. The merger orbits for the matched pairs in the TNG100-1 and the TNG100-1-Dark simulations are similar right after infall, and both evolve to more head-on orbits at final stages, with smaller changes in the hydrodynamical simulation. In the final 2 Gyr before merger, the collision angles that represent merger orbits quantitatively are smaller in TNG100-1 than those in TNG100-1-Dark, by around 6$^\circ$ to 10$^\circ$, depending on the mass ratios and galaxy masses investigated. Our results demonstrate that the presence of baryons accelerates a bit the merger processes, and results in more spiral-in orbits for both major and minor mergers in galaxies with various stellar masses. These effects are less obvious in simulations with lower resolutions.

astro-ph.GA↗