Searcharxiv⌕ Search

arXiv subjects

Ben Wang

Publications and source records attributed to Ben Wang.

At least 19 recordsLinked to original sources
arXiv2026

FinProBench: Evaluating Financial AI Agents with Role-Grounded Rubrics Derived from Professional Deliverables

Evaluating financial AI agents requires criteria aligned with real professional work. Existing rubric methods typically derive criteria from task prompts or model outputs, overlooking tacit standards visible only in practitioner deliverables. We introduce FinProBench, a benchmark for professional financial tasks, and Role-Grounded Rubric Construction (RGRC), a reusable pipeline that derives rubrics from deliverables produced by practitioners in the same role. RGRC comprises four stages: Deliverable Collection, Competency Extraction, Rubric Synthesis, and Validation. Its rubrics capture tacit standards, distinguish quality levels, and transfer across tasks within a role. Before analysis, we classified 57 occupations by deliverable genre into 30 prior-rich conventional roles and 27 prior-sparse role-specialized roles. Across all roles, Prompt-only nearly matches RGRC for conventional roles (89.2% vs. 90.7%), but RGRC substantially outperforms it for role-specialized roles (99.1% vs. 78.0%). This split indicates that prompt engineering can approximate rubrics when conventions are well represented in model priors, while professional grounding is essential for standards beyond those priors. FinProBench is built from 1,723 curated deliverables spanning 57 occupations, 8 financial sub-industries, and 161 deliverable types, and releases an initial evaluation set of 20 complete tasks covering 20 roles in 7 sub-industries. With heterogeneous LLM judges and role-level rubrics, human deliverables rank first on average (73.7 vs. 70.3, 70.2, and 69.6 out of 100), while all four systems show overlapping 95% confidence intervals and complementary strengths. Reusing rubrics at the role level reduces estimated per-task construction effort by 6.7 times relative to authoring each rubric from scratch.

cs.AI↗
arXiv2026

FinPerMA: A Theory-Informed, Event-Grounded Personalized-Memory Benchmark for LLM Agents

Large language model (LLM) agents are increasingly used as personalized assistants in high-stakes domains such as financial advising, yet it remains unclear whether they can maintain and update an individualized user model over long horizons. Existing personalized-memory benchmarks primarily test factual retention or rely on weakly constrained model-generated trajectories, leaving event-driven preference adaptation underexplored. We introduce FinPerMA, an event-grounded benchmark that evaluates personalized memory against frozen longitudinal investor trajectories. Its generation pipeline combines deterministic, theory-informed impact rules, controlled LLM narration, and automated quality screening; a Post-Shock checkpoint isolates whether an agent has integrated a material event into its persistent user model. On 2,994 questions from 276 personas, seven frontier LLMs and up to seven memory configurations remain far from saturated: no full-context configuration exceeds approximately 0.47 overall accuracy or approximately 39% on multiple-choice questions. Attribution analysis shows that summary-based memory often preserves factual details while losing the preference signals needed for personalization; simple retrieval can therefore outperform purpose-built memory systems, with the gap widening after shocks.

cs.AI↗
arXiv2026

Dose-efficient Quantum Phase Estimation in Lossy Optical Interferometry

Optical interferometry is a cornerstone technique for precise phase measurements across various fields. In many applications, for example, biological imaging, it often necessitates stringent limits on light intensity to prevent adverse effects on light-sensitive samples, a condition known as dose-limited regimes. Maximizing the precision per dose is therefore crucial. In quantum metrology, quantum correlations enable high precision in phase estimation while adhering to dose constraints. Nevertheless, photon loss, including absorption by a sample, substantially diminishes the benefits of quantum enhancement in interferometry. In this work, we experimentally investigate a dose-efficient approach to quantum phase estimation using sequential strategies in the presence of loss. Performance of sequential strategies with and without control is evaluated through quantum Fisher information (QFI) per dose. Experimental results show that both sequential strategies exceed the classical limit and outperform the parallel strategy using unbalanced N00N states. Notably, the control-enhanced sequential strategy attains superior QFI per dose, approaching the quantum limit. These results highlight the promise of sequential strategy for imaging and sensing in resource-constrained scenarios, marking a significant step toward practical and efficient quantum metrology in lossy environments.

quant-ph↗
arXiv2026

BilliardPhys-Bench: Benchmarking Physical Reasoning and Visual Dynamics of Multimodal LLMs

Current multimodal models handle static image recognition well, but intuitive physical reasoning remains a weakness. Predicting how objects will move and interact from a single image is still difficult for these systems. We present BilliardPhys-Bench, a benchmark for physical reasoning in synthetic billiards environments. Its procedural engine generates randomized scenarios with friction and elastic collisions. The benchmark tests three abilities: (1) predicting ball-to-ball collisions, (2) reasoning about wall bounces, and (3) estimating final ball positions after motion stops. We evaluate recent MLLMs from the GPT, Claude, Gemini, and Qwen families. Performance drops as simulation time increases and scene geometry grows more complex. We also observe a consistent failure mode we call "stasis bias": when the correct physical outcome is harder to infer, models tend to predict no interaction. These findings show where current MLLMs break down on visual dynamics and point toward the need for better physical inductive biases in multimodal architectures.

cs.AI↗
arXiv2026

Constraining the nature of active galactic nuclei through circumgalactic Lya emission at z=2-3

We present a comprehensive analysis of circumgalactic Lya nebulae around 59 unobscured and 26 obscured quasars at z=2-3, observed with the Keck Cosmic Web Imager (KCWI), to constrain the nature of active galactic nuclei (AGN) at cosmic noon. We find that Lya nebulae around unobscured quasars are significantly less symmetric having a symmetry parameter of a_w=0.2-0.6 and more spatially extended having a scale length of r_h=10.7+/-0.5 kpc than those around obscured quasars (a_w=0.6-0.8; r_h=6.6-7.7 kpc).Unobscured quasars also exhibit steeply declining velocity dispersion profiles with the slope of -4.3+/-0.4 km s^-1 kpc^-1, indicative of large-scale outflows, whereas obscured quasars display flat profiles (-0.2+/-0.7 and -0.6+/-0.4 km s^-1 kpc^-1). The degree of quasar obscuration appears to be intrinsically linked to nebular asymmetry and extent, a relationship that could be in tension with the standard orientation-based AGN unified model, as it expects unobscured-quasar nebulae to be more symmetric and compact. These results naturally fit the evolutionary scenario, where AGN feedback drives a transition from an obscured to an unobscured phase-progressively redistributing gas to larger radii, introducing anisotropy, and driving turbulence. Taken together, our findings favor the evolutionary scenario over the purely orientation-based unified model for quasars at cosmic noon.

astro-ph.GA↗
arXiv2026

FeynmanBench: Benchmarking Multimodal LLMs on Diagrammatic Physics Reasoning

Current multimodal benchmarks for scientific reasoning primarily evaluate local information extraction -- models recognize symbols and values and then perform textual inference. They do not assess whether models can reason over the global structural properties of formal diagrams, such as topology, conservation constraints, and the consistent mapping between visual patterns and algebraic expressions. We introduce FeynmanBench, a benchmark of over 2,000 tasks centered on Feynman diagrams spanning the electromagnetic, weak, and strong interactions of the Standard Model. Each instance couples a diagram image with minimal textual conventions and requires models to recover the full physical content -- vertex inventory, propagator types, topological connectivity, momentum routing, and the complete scattering amplitude. An automated generation and verification pipeline produces the diagrams, annotations, and reference answers under standardized rules. Evaluating 19 state-of-the-art multimodal LLMs, we find a consistent failure pattern: models achieve 70--95\% on local recognition (vertex and propagator identification) but collapse to 13--17\% on topological reconstruction (CP3), and near zero on full algebraic derivation (CP5). FeynmanBench offers a controlled testbed for multimodal reasoning over formal scientific diagrams and highlights fundamental limitations of current architectures in topology-sensitive scientific reasoning.

cs.AI↗
arXiv2026

ResAdapt: Adaptive Resolution for Efficient Multimodal Reasoning

Multimodal Large Language Models (MLLMs) achieve stronger visual understanding by scaling input fidelity, yet the resulting visual token growth makes jointly sustaining high spatial resolution and long temporal context prohibitive. We argue that the bottleneck lies not in how post-encoding representations are compressed but in the volume of pixels the encoder receives, and address it with ResAdapt, an Input-side adaptation framework that learns how much visual budget each frame should receive before encoding. ResAdapt couples a lightweight Allocator with an unchanged MLLM backbone, so the backbone retains its native visual-token interface while receiving an operator-transformed input. We formulate allocation as a contextual bandit and train the Allocator with Cost-Aware Policy Optimization (CAPO), which converts sparse rollout feedback into a stable accuracy-cost learning signal. Across budget-controlled video QA, temporal grounding, and image reasoning tasks, ResAdapt improves low-budget operating points and often lies on or near the efficiency-accuracy frontier, with the clearest gains on reasoning-intensive benchmarks under aggressive compression. Notably, ResAdapt supports up to 16x more frames at the same visual budget while delivering over 15% performance gain. Code is available at https://github.com/Xnhyacinth/ResAdapt.

cs.CV↗
arXiv2026

SPM-Bench: Benchmarking Large Language Models for Scanning Probe Microscopy

As LLMs achieved breakthroughs in general reasoning, their proficiency in specialized scientific domains reveals pronounced gaps in existing benchmarks due to data contamination, insufficient complexity, and prohibitive human labor costs. Here we present SPM-Bench, an original, PhD-level multimodal benchmark specifically designed for scanning probe microscopy (SPM). We propose a fully automated data synthesis pipeline that ensures both high authority and low-cost. By employing Anchor-Gated Sieve (AGS) technology, we efficiently extract high-value image-text pairs from arXiv and journal papers published between 2023 and 2025. Through a hybrid cloud-local architecture where VLMs return only spatial coordinates "llbox" for local high-fidelity cropping, our pipeline achieves extreme token savings while maintaining high dataset purity. To accurately and objectively evaluate the performance of the LLMs, we introduce the Strict Imperfection Penalty F1 (SIP-F1) score. This metric not only establishes a rigorous capability hierarchy but also, for the first time, quantifies model "personalities" (Conservative, Aggressive, Gambler, or Wise). By correlating these results with model-reported confidence and perceived difficulty, we expose the true reasoning boundaries of current AI in complex physical scenarios. These insights establish SPM-Bench as a generalizable paradigm for automated scientific data synthesis.

cs.AI↗
arXiv2026

HLE-Verified: A Systematic Verification and Structured Revision of Humanity's Last Exam

Humanity's Last Exam (HLE) has become a widely used benchmark for evaluating frontier large language models on challenging, multi-domain questions. However, community-led analyses have raised concerns that HLE contains a non-trivial number of noisy items, which can bias evaluation results and distort cross-model comparisons. To address this challenge, we introduce HLE-Verified, a verified and revised version of HLE with a transparent verification protocol and fine-grained error taxonomy. Our construction follows a two-stage validation-and-repair workflow resulting in a certified benchmark. In Stage I, each item undergoes binary validation of the problem and final answer through domain-expert review and model-based cross-checks, yielding 668 verified items. In Stage II, flawed but fixable items are revised under strict constraints preserving the original evaluation intent, through dual independent expert repairs, model-assisted auditing, and final adjudication, resulting in 1,143 revised-and-certified items. The remaining 689 items are released as a documented uncertain set with explicit uncertainty sources and expertise tags for future refinement. We evaluate eight state-of-the-art language models on HLE and HLE-Verified, observing an average absolute accuracy gain of 7--10 percentage points on HLE-Verified. The improvement is particularly pronounced on items where the original problem statement and/or reference answer is erroneous, with gains of 30--40 percentage points. Our analyses further reveal a strong association between model confidence and the presence of errors in the problem statement or reference answer, supporting the effectiveness of our revisions. Overall, HLE-Verified improves HLE-style evaluations by reducing annotation noise and enabling more faithful measurement of model capabilities. Data is available at: https://huggingface.co/datasets/skylenage/HLE-Verified

cs.CL↗
arXiv2026

MMSF: Multitask and Multimodal Supervised Framework for WSI Classification and Survival Analysis

Multimodal evidence is critical in computational pathology: gigapixel whole slide images capture tumor morphology, while patient-level clinical descriptors preserve complementary context for prognosis. Integrating such heterogeneous signals remains challenging because feature spaces exhibit distinct statistics and scales. We introduce MMSF, a multitask and multimodal supervised framework built on a linear-complexity MIL backbone that explicitly decomposes and fuses cross-modal information. MMSF comprises a graph feature extraction module embedding tissue topology at the patch level, a clinical data embedding module standardizing patient attributes, a feature fusion module aligning modality-shared and modality-specific representations, and a Mamba-based MIL encoder with multitask prediction heads. Experiments on CAMELYON16 and TCGA-NSCLC demonstrate 2.1--6.6\% accuracy and 2.2--6.9\% AUC improvements over competitive baselines, while evaluations on five TCGA survival cohorts yield 7.1--9.8\% C-index improvements compared with unimodal methods and 5.6--7.1\% over multimodal alternatives.

cs.CV↗
arXiv2026

The Geometric Reasoner: Manifold-Informed Latent Foresight Search for Long-Context Reasoning

Scaling test-time compute enhances long chain-of-thought (CoT) reasoning, yet existing approaches face a fundamental trade-off between computational cost and coverage quality: either incurring high training expense or yielding redundant trajectories. We introduce The Geometric Reasoner (TGR), a training-free framework that performs manifold-informed latent foresight search under strict memory bounds. At each chunk boundary, TGR scores candidate latent anchors via a lightweight look-ahead estimate combined with soft geometric regularizers that encourage smooth trajectories and diverse exploration. Chunk-wise KV cache resets keep memory linear in chunk length. On challenging math and code benchmarks, TGR improves robust trajectory coverage, measured by the area under the Pass@k curve (AUC), by up to 13 points on Qwen3-8B, with negligible overhead of about 1.1--1.3 times.

cs.LG↗
arXiv2025

Experimental Joint Estimation of Phase and Phase Diffusion via Deterministic Bell Measurements

Accurate phase estimation plays a pivotal role in quantum metrology, yet its precision is significantly affected by noise, particularly phase-diffusive noise caused by phase drift. To address this challenge, the joint estimation of phase and phase diffusion has emerged as an effective approach, transforming the problem into a multi-parameter estimation task. However, the incompatibility between optimal measurements for different parameters prevents single-copy measurements from reaching the fundamental precision limits defined by the quantum Cramer-Rao bound. Meanwhile, collective measurements performed on multiple identical copies can mitigate this incompatibility and thus enhance the precision of joint parameter estimation. This work experimentally demonstrates joint phase and phase-diffusion estimation using deterministic Bell measurements on a two-qubit system. A linear optical network is employed to implement both parameter encoding and deterministic Bell measurements, achieving improved estimation precision compared to any separable measurement strategy. This work proposes a new framework for phase estimation under phase-diffusive noise and underscores the substantial advantages of collective measurements in multi-parameter quantum metrology.

quant-ph↗
arXiv2025

Your plan may succeed, but what about failure? Investigating how people use ChatGPT for long-term life task planning

Long-term life task planning is inherently complex and uncertain, yet little is known about how emerging AI systems support this process. This study investigates how people use ChatGPT for such planning tasks, focusing on user practices, uncertainties, and perceptions of AI assistance. We conducted an interview study with 14 participants who engaged in long-term planning activities using ChatGPT, combining analysis of their prompts and interview responses. The task topics across diverse domains, including personal well-being, event planning, and professional learning, along with prompts to initiate, refine, and contextualize plans. ChatGPT helped structure complex goals into manageable steps, generate ideas, and sustain motivation, serving as a reflective partner. Yet its outputs were often generic or idealized, lacking personalization, contextual realism, and adaptability, requiring users to actively adapt and verify results. Participants expressed a need for AI systems that provide adaptive and trustworthy guidance while acknowledging uncertainty and potential failure in long-term planning. Our findings show how AI supports long-term life task planning under evolving uncertainty and highlight design implications for systems that are adaptive, uncertainty-aware, and capable of supporting long-term planning as an evolving human-AI collaboration.

cs.HC↗
arXiv2025

Solar PV Installation Potential Assessment on Building Facades Based on Vision and Language Foundation Models

Building facades represent a significant untapped resource for solar energy generation in dense urban environments, yet assessing their photovoltaic (PV) potential remains challenging due to complex geometries and semantic com ponents. This study introduces SF-SPA (Semantic Facade Solar-PV Assessment), an automated framework that transforms street-view photographs into quantitative PV deployment assessments. The approach combines com puter vision and artificial intelligence techniques to address three key challenges: perspective distortion correction, semantic understanding of facade elements, and spatial reasoning for PV layout optimization. Our four-stage pipeline processes images through geometric rectification, zero-shot semantic segmentation, Large Language Model (LLM) guided spatial reasoning, and energy simulation. Validation across 80 buildings in four countries demonstrates ro bust performance with mean area estimation errors of 6.2% ± 2.8% compared to expert annotations. The auto mated assessment requires approximately 100 seconds per building, a substantial gain in efficiency over manual methods. Simulated energy yield predictions confirm the method's reliability and applicability for regional poten tial studies, urban energy planning, and building-integrated photovoltaic (BIPV) deployment. Code is available at: https:github.com/CodeAXu/Solar-PV-Installation

cs.CV↗
arXiv2025

EfficientMIL: Efficient Linear-Complexity MIL Method for WSI Classification

Whole slide images (WSIs) classification represents a fundamental challenge in computational pathology, where multiple instance learning (MIL) has emerged as the dominant paradigm. Current state-of-the-art (SOTA) MIL methods rely on attention mechanisms, achieving good performance but requiring substantial computational resources due to quadratic complexity when processing hundreds of thousands of patches. To address this computational bottleneck, we introduce EfficientMIL, a novel linear-complexity MIL approach for WSIs classification with the patches selection module Adaptive Patch Selector (APS) that we designed, replacing the quadratic-complexity self-attention mechanisms in Transformer-based MIL methods with efficient sequence models including RNN-based GRU, LSTM, and State Space Model (SSM) Mamba. EfficientMIL achieves significant computational efficiency improvements while outperforming other MIL methods across multiple histopathology datasets. On TCGA-Lung dataset, EfficientMIL-Mamba achieved AUC of 0.976 and accuracy of 0.933, while on CAMELYON16 dataset, EfficientMIL-GRU achieved AUC of 0.990 and accuracy of 0.975, surpassing previous state-of-the-art methods. Extensive experiments demonstrate that APS is also more effective for patches selection than conventional selection strategies.

cs.CV↗
arXiv2025

Superresolution of unequal-brightness thermal sources for stellar interferometry

Resolving high-contrast targets is a fundamental yet highly challenging task in astronomy. Using quantum estimation theory, we demonstrate that the ultimate limit for estimating the separation between two unequal-brightness thermal sources via interferometry remains constant, enabling the potential for superresolution. We give a comparative analysis of two primary stellar interferometric schemes: amplitude interferometry and intensity interferometry. Notably, the nulling strategy employed in amplitude interferometry, a configuration specifically proposed for exoplanet detection by leveraging destructive interference to suppress the brighter source, is quantum optimal for separation estimation. While intensity interferometry is less effective than amplitude interferometry in lossless scenarios and fails to achieve superresolution, it becomes competitive when optical loss in large-scale interferometry is considered. By applying these methodologies to modern stellar interferometry, we highlight the promise of large-scale interferometry for advancing high-resolution astronomical observation.

astro-ph.IM↗
arXiv2025

Approaching the Multiparameter Quantum Cram\'er-Rao Bound via Classical Correlation and Entangling Measurements

Multiparameter quantum metrology is essential for a wide range of practical applications. However, simultaneously achieving the ultimate precision for all parameters, as prescribed by the quantum Cram\'er-Rao bound (QCRB), remains a significant challenge. In this work, we propose a scheme termed local operation with entangling measurements (LOEM) strategy, which leverages classically correlated orthogonal pure states combined with entangling measurements to attain the multiparameter QCRB. We experimentally validate this scheme using a quantum photonic system. Additionally, we employ iterative interactions to demonstrate that the LOEM strategy can achieve the precision of Heisenberg scaling. By theoretically and experimentally demonstrating the saturation of the multiparameter QCRB with the LOEM strategy, our work advances the practical applications of quantum metrology in multiparameter estimation.

quant-ph↗
arXiv2025

Luminous Mid-IR Selected Obscured Quasars at Cosmic Noon in SDSS Stripe82 II: Spectroscopic Diversity and Broad H$\alpha$ Emissions

We present a multiwavelength spectroscopic survey of 23 luminous mid-infrared-selected Type-2 quasars at redshifts z = 0.88 to 3.49. The targets were selected in the SDSS Stripe 82 field based on their bright WISE W4 detections (flux > 5 mJy) and extremely faint or red optical counterparts (e.g., r > 23 or r - W4 > 8.4), designed to identify heavily obscured quasars. Deep near-infrared (Gemini/GNIRS) and optical (Keck/LRIS and KCWI) spectroscopy confirm 23 out of 24 candidates as Type-2 quasars in this redshift range, including 12 objects at z > 2. The spectra exhibit strong rest-frame UV and optical emission lines (Ly-alpha, C IV, [O III], H-alpha) with a wide range of line widths, indicating significant spectral diversity. Approximately one-third of the sample (8 of 23) shows broad H-alpha emission (FWHM > 2000 km/s) despite their Type-2 classification, while the rest have only narrow lines (FWHM < 2000 km/s) characteristic of classical obscured quasars. Notably, these broad-line Type-2 quasars share similar spectral energy distributions with the JWST-discovered "little red dot" (LRD) AGNs, suggesting that our sample could be lower-redshift analogues of the heavily obscured broad-line AGNs uncovered by JWST. We also find that the [O III] 5007 angstrom emission is relatively weak for their high bolometric luminosities, deviating from trends seen in lower-redshift Type-2 QSOs. A new composite spectrum for Type-2 QSOs is built using our sample. Overall, our results demonstrate that mid-IR selection efficiently uncovers a diverse population of obscured quasars and that spectroscopic follow-up is crucial for revealing their true nature. This study provides new insights into heavily obscured SMBH growth at cosmic noon and bridges the gap to the obscured AGN populations now being revealed by JWST.

astro-ph.GA↗