Search arXiv⌕ Search

arXiv subjects

Ran Bi

Publications and source records attributed to Ran Bi.

At least 19 recordsLinked to original sources
arXiv2026

Ask on Demand: Targeted Peer Communication for Parallel Agentic Systems

Parallel LLM agents tackle complex tasks by exploring multiple trajectories, yet useful discoveries remain isolated in individual execution histories. Existing sharing methods often select or summarize information before other agents' needs become clear, potentially overlooking details that become important later. We propose Ask on Demand (AOD), a communication framework that enables agents to seek targeted peer assistance when specific information needs arise. Upon an approved request, AOD retrieves relevant evidence from peer execution histories, selects a suitable peer, and generates a response tailored to the request. To determine when such assistance is worthwhile, we train a lightweight gate using paired team continuations, with supervision derived from final team outcomes and inference cost. Experiments on AppWorld and WebShop demonstrate that AOD consistently improves task performance over independent parallel execution while achieving lower inference cost than existing information-sharing baselines. These findings highlight the benefits of aligning peer communication with agents' evolving information needs to achieve more effective and efficient collaboration.

cs.AI↗
arXiv2026

CityCode: Building Agent-Operable 3D Cities from Sparse Aerial Observations

Urban spatial intelligence calls for 3D cities that agents can measure, interpret, and modify across successive tasks. Maintaining such cities as persistent, code-based states requires lightweight acquisition and support for consistent state updates without repeated dense reconstruction. We therefore formulate a new setting for grounded urban world construction: generating large-scale, instance-level parameterized 3D cities from one low-altitude aerial RGB image per region and known camera parameters, while recovering metrically scaled buildings and explicit road geometry and connectivity. We introduce CityCode, an agent-operable representation that unifies metric geometry, urban topology, visual semantics, and 3D assets through persistent entity identities, with dependency-aware versioned execution for validated updates and historical restoration. We further develop Air2Code to construct this representation by combining visual structure extraction with adapted metric depth estimation. To address occlusion and incomplete coverage, recovered structure and observed appearance guide generative completion and subsequent structure-aware 3D generation for buildings and backgrounds. Experiments on CitySample-Aerial and UrbanScene3D demonstrate substantially more accurate metric structure recovery and superior visual fidelity to the target cities compared with the evaluated baselines. Across 50 task chains of 50 steps each, CityCode supports continued querying, editing, planning, and historical restoration over evolving city states. Together, CityCode and Air2Code turn sparse aerial observations into persistent, agent-operable urban worlds for long-horizon urban spatial intelligence.

cs.CV↗
arXiv2026

Fully Deterministic Shallow Neural Networks with Localized Activations. Part I: Construction and Optimal Sobolev Approximation

We construct shallow neural networks with fixed Gaussian and tanh-difference activations on bounded Lipschitz domains $Ω\subset\mathbb{R}^d$, $d\ge2$. For each $s>0$, explicit deterministic rules prescribe all hidden parameters independently of the target $u\in H^s(Ω)$. At frequency resolution $M$, the $N_M\asymp M^d$ prescribed features span a linear space $V_M$. For every sufficiently large integer $N$, we choose $M\asymp N^{1/d}$ with $N_M\le N$ and construct a bounded linear operator $A_N:H^s(Ω)\to V_M$ such that \[ \|u-A_Nu\|_{H^m(Ω)} \le C N^{-(s-m)/d}\|u\|_{H^s(Ω)} \] for every $u\in H^s(Ω)$ and every integer $0\le m<s$, with $C$ independent of $u$, $N$, and $m$. Sobolev width lower bounds show that this rate is optimal among linear spaces of dimension at most $N$ chosen independently of the target. We use explicit spherical cubature nodes for the directions and Gauss-Legendre nodes or equispaced midpoints for the offsets. We show that the exterior offset margin can shrink to zero as $M\to\infty$ while preserving the approximation rate. A margin beyond $\sup_{x\inΩ}|x|$ of order $M^{-1}\log M$ for the Gaussian and $M^{-1}(\log M)^2$ for the tanh difference suffices to truncate the ridge integral and remove exterior midpoint nodes. Numerical experiments in two and three dimensions examine $L^2$ and $H^1$ approximation.

math.NA↗
arXiv2026

Relevance Is Not Sufficiency: What Actually Closes the Evidence Gap in Long-Term Memory QA

LLM agents that interact with a user across many sessions accumulate histories that exceed their context window, so they store past interactions in an external memory and answer each question from a small set of retrieved records. Existing memory systems rank records by lexical or embedding relevance, yet the top-ranked memories can each be relevant while jointly omitting a complementary fact that the answer requires, especially for multi-session and temporal questions. Drawing on the distinction between relevance and sufficiency in legal evidence scholarship, we recast memory retrieval as constructing a sufficient memory set. To operationalize this view, we introduce a blinded LLM judgment over the retrieved set, together with Gold Hit and Turn Hit as evidence-coverage proxies. We then propose Budgeted Flat Reconstruction (BFR), which builds sufficient sets over a fixed flat memory store in two stages. Specifically, we first apply Formal Concept Analysis for Memory Selection (FCA-MS) to decompose the question into information requirements and select a compact candidate subset that jointly covers them. Then, we repeatedly acquire unseen records through deeper text search or complementary entity and session views, stopping when the budget is exhausted. Experiments on LoCoMo and LongMemEval-S show that BFR outperforms same-store adaptations of recent agent-memory systems in both answer quality and evidence coverage. Specifically, on LongMemEval-S it raises judged accuracy from 72.4% to 82.2% and Turn Hit to 91.4%.

cs.AI↗
arXiv2026

Fully Deterministic Shallow Neural Networks with Localized Activations. Part II: Condition Number Estimates

We prove positive definiteness and quantitative condition-number bounds for continuous Sobolev Gram matrices of deterministic shallow networks with Gaussian and tanh-difference activations. The prescribed feature families include the optimal Sobolev approximation constructions of Part I. On every fixed nonempty bounded Lipschitz domain in $\R^d$, $d\ge2$, and for every fixed integer $m\ge0$, the raw and column-normalized Gram matrices in the full $H^m$ inner product have spectral condition numbers bounded by $\exp(CN^{2/d}/\log(N+2))$ for sufficiently large $N$, where $N$ is the feature count and $C>0$ is independent of $N$. The upper bounds hold for midpoint and Gauss-Legendre offsets. For midpoint offsets and a suitable angular sequence, matching lower bounds on fixed ellipsoids establish sharpness of this exponent over the stated domain class for raw Gaussian matrices and both raw and normalized tanh-difference matrices. The proof combines homogeneous differential operators that isolate direction groups with one-dimensional coefficient-recovery estimates. Two-dimensional $L^2$ experiments on a square illustrate condition-number growth and normalization effects.

math.NA↗
arXiv2026

A Two-stage Adaptive Lifting PINN Framework for Solving Viscous Approximations to Hyperbolic Conservation Laws

Training physics-informed neural networks (PINNs) for hyperbolic conservation laws near the inviscid limit is challenging because shock discontinuities invalidate classical strong-form residuals, while small-viscosity regularization creates narrow internal layers that aggravate spectral bias. We propose a Two-stage Adaptive Lifting PINN (TAL-PINN), a fully autonomous framework that augments the physical coordinates with a learned auxiliary field constructed through $r$-adaptive coordinate transformations. The auxiliary geometry is extracted automatically from a computationally inexpensive high-viscosity coarse solution, providing a pre-conditioning mechanism that guides the network toward steep-gradient regions without prior knowledge of shock dynamics. The coordinate construction and lifted training can also be embedded in a viscosity-continuation procedure. Theoretically, we derive a scalar a posteriori $L^2$ error estimate that reveals the viscosity dependence of residual-based error control, establish a variance-reduction criterion for coordinate-induced importance sampling, and use neural tangent kernel analysis to explain how adaptive augmentation enriches tangent features and accelerates residual decay. Numerical experiments on scalar equations and the Euler system show faster loss decay and improved small-viscosity accuracy. Ablation studies further show that lifting yields additional gains beyond viscosity continuation and adaptive sampling alone.

math.NA↗
arXiv2026

Decreasing Digital Distraction in College Students: Associated Online Learning Strategies Identified by Unsupervised Data Mining Approaches

The proliferation of digital tools in education offers numerous benefits but also introduces significant challenges, notably digital distractions that hinder academic performance, especially in online learning contexts. This study employed unsupervised data mining techniques, specifically association rule mining and clustering analysis, to identify effective learning strategies associated with lower levels of digital distractions among college students. Data from 530 participants revealed that self-regulated learning strategies (i.e., goal setting, environment structuring, and time management) co-occurred most consistently with lower digital distractions. Additionally, learner-instructor and learner-content engagement strategies, as well as technical competencies, also tended to appear in the same profiles as lower distraction. Interestingly, reliance on peer help-seeking and learner-learner engagement strategies appeared less often in those lower distraction profiles. These findings offer actionable implications for educators to design targeted interventions that foster focused and productive online learning environments.

cs.CY↗
arXiv2026

Adaptive Randomized Neural Networks with Locally Activation Function: Theory and Algorithm for Solving PDEs

This paper establishes an approximation theory and develop an adaptive computational framework for randomized neural networks (RaNNs). For RaNNs of the form $\sum_{i=1}^{N} W_i σ(A_i\cdot x+B_i)$ with hidden parameters uniformly sampled from a bounded set of scale $M$, we show that the choice $M \asymp N^{p/[2((p-1)(d+1)+pη)]}$ yields an expected $W^{k,p}$-approximation error of order $\mathcal{O}\left(N^{-pη/[2((p-1)(d+1)+pη)]}\right)$, where $η$ is associated with the regularity of the target function in a generalized Barron spectral space. This theoretical result demonstrates that lower regularity requires a larger sampling range for the hidden parameters. Motivated by the relationship between $M$ and $η$, we introduce an adaptive physics-informed RaNN (PIRaNN) method that which couples parameter sampling with an adaptive partition of unity via local affine scaling. Residual-based a posteriori error indicators and Dörfler marking are used to drive local refinement. Numerical experiments, including the 1D viscous Burgers' equation and 2D/3D problems with localized peaks and L-shaped corner singularities, demonstrate that our method preferentially refines regions exhibiting large gradients and singularities. Compared to standard non-adaptive RaNNs, the adaptive PIRaNN effectively captures complex local features with significantly enhanced accuracy and efficiency.

math.NA↗
arXiv2026

Does Discussion Matter? Interaction Dynamics in Transparent Peer Review

Open peer review is advocated as a way to make review more dialogic, yet evidence on whether it changes participant behavior remains scarce. We compare two transparent review designs: Nature Human Behaviour, an editor-mediated model, and International Conference on Learning Representations 2023, a discussion-based model. Treating cross-mentions, moments when one participant explicitly names another beyond the immediate dyad, as the analytic unit, we combine social network analysis of 56 NHB articles and 371 ICLR notable papers with qualitative coding grounded in Knowledge Building theory. Structurally, NHB forms hub-and-spoke networks with zero reviewer-to-reviewer reciprocity, whereas ICLR forms partial-mesh networks with significantly more direct cross-referencing. Functionally, the distribution of communicative functions is similar across venues, but who performs them differs: editors aggregate at NHB, while authors and reviewers absorb that labor at ICLR. The designs differ most in how communicative labor is distributed across roles, suggesting platforms should support it as explicit, built-in functions.

cs.HC↗
arXiv2026

Prober.ai: Gated Inquiry-Based Feedback via LLM-Constrained Personas for Argumentative Writing Development

The proliferation of large language models (LLMs) in educational settings has paradoxically undermined the cognitive processes they purport to support. Students increasingly outsource critical thinking to AI assistants that generate polished text on demand, resulting in measurable cognitive debt and diminished argumentative reasoning skills. We present Prober.ai, a web-based writing environment that inverts the conventional AI-tutoring paradigm: rather than generating or rewriting student text, the system constrains an LLM (Gemini 3 Flash Preview) through persona-specific system prompts and structured JSON output schemas to produce only targeted, inquiry-based questions about argumentative weaknesses. A two-phase interaction architecture -- Challenge and Unlock -- implements a pedagogical friction mechanism whereby revision suggestions are gated behind mandatory student reflection. The system's design is grounded in Toulmin's argumentation theory, research on peer feedforward questioning mechanisms, and evidence on AI-supported feedback in writing instruction. A functional prototype was developed in 36 hours during the NY EdTech Hackathon (March 2026), where it was awarded second place. We describe the system architecture, the prompt engineering methodology for constraining LLM output to pedagogically aligned JSON schemas, and discuss implications for scalable, cognition-preserving AI integration in writing education.

cs.AI↗
arXiv2026

Stability Annealing Selects the Implicit Bias of Smoothed Sign Descent: A Rate-Indexed Barrier Path on Separable Data

Adaptive gradient methods can favor max-margin separators that differ from gradient descent, yet a fixed positive numerical stability constant eventually changes the update geometry again. This paper studies the rate-controlled middle case for full-batch linear classification on separable data. For memoryless stability-annealed smoothed-sign descent with weighted exponential loss, we prove that the normalized iterates converge to the minimizer of a convex Burg-type barrier over a margin slice. The proof rewrites the dynamics exactly as entropic mirror ascent on a concave dual objective, controls the dual gap by a KL recursion, and yields an explicit S_t^{-1/2} normalized-iterate envelope. The static barrier geometry is fully characterized, including KKT conditions and both endpoint limits. Experiments validate the exact dual identities to floating-point error, illustrate the predicted path and rate diagram, and show an empirical fixed-epsilon crossover scaling in cumulative time. We further report robustness and boundary diagnostics for logistic tails, fixed-epsilon crossover, and adaptive-method variants, delineating the scope of the proved smoothed-sign theory.

cs.LG↗
arXiv2026

Extended Interface Physics-Informed Neural Networks Method for Moving Interface Problems

Physics-informed neural networks (PINNs) have emerged as an effective class of mesh-free methods for solving partial differential equations (PDEs), particularly on complex geometries. In this paper, we introduce an Extended Interface Physics-Informed Neural Network (XI-PINN) framework designed to solve parabolic moving interface problems. The proposed method employs a level set function--which can be either analytically prescribed or learned via a neural network--to capture the moving interface. Furthermore, we establish an a priori error analysis for the XI-PINN method and derive error bounds for the approximation. Extensive numerical experiments are provided to validate the accuracy and robustness of the framework, and its applicability is further demonstrated by solving the Oseen equations.

math.NA↗
arXiv2025

Towards Seeing Bones at Radio Frequency

Wireless sensing literature has long aspired to achieve X-ray-like vision at radio frequencies. Yet, state-of-the-art wireless sensing literature has yet to generate the archetypal X-ray image: one of the bones beneath flesh. In this paper, we explore MCT, a penetration-based RF-imaging system for imaging bones at mm-resolution, one that significantly exceeds prior penetration-based RF imaging literature. Indeed the long wavelength, significant attenuation and complex diffraction that occur as RF propagates through flesh, have long limited imaging resolution (to several centimeters at best). We address these concerns through a novel penetration-based synthetic aperture algorithm, coupled with a learning-based pipeline to correct for diffraction-induced artifacts. A detailed evaluation of meat models demonstrates a resolution improvement from sub-decimeter to sub-centimeter over prior art in RF penetrative imaging.

cs.GR↗
arXiv2025

UniGenX: a unified generative foundation model that couples sequence, structure and function to accelerate scientific design across proteins, molecules and materials

Function in natural systems arises from one-dimensional sequences forming three-dimensional structures with specific properties. However, current generative models suffer from critical limitations: training objectives seldom target function directly, discrete sequences and continuous coordinates are optimized in isolation, and conformational ensembles are under-modeled. We present UniGenX, a unified generative foundation model that addresses these gaps by co-generating sequences and coordinates under direct functional and property objectives across proteins, molecules, and materials. UniGenX represents heterogeneous inputs as a mixed stream of symbolic and numeric tokens, where a decoder-only autoregressive transformer provides global context and a conditional diffusion head generates numeric fields steered by task-specific tokens. Besides the new high SOTAs on structure prediction tasks, the model demonstrates state-of-the-art or competitive performance for the function-aware generation across domains: in materials, it achieves "conflicted" multi-property conditional generation, yielding 436 crystal candidates meeting triple constraints, including 11 with novel compositions; in chemistry, it sets new benchmarks on five property targets and conformer ensemble generation on GEOM; and in biology, it improves success in modeling protein induced fit (RMSD < 2 Å) by over 23-fold and enhances EC-conditioned enzyme design. Ablation studies and cross-domain transfer substantiate the benefits of joint discrete-continuous training, establishing UniGenX as a significant advance from prediction to controllable, function-aware generation.

cs.LG↗
arXiv2025

Nature Language Model: Deciphering the Language of Nature for Scientific Discovery

Foundation models have revolutionized natural language processing and artificial intelligence, significantly enhancing how machines comprehend and generate human languages. Inspired by the success of these foundation models, researchers have developed foundation models for individual scientific domains, including small molecules, materials, proteins, DNA, RNA and even cells. However, these models are typically trained in isolation, lacking the ability to integrate across different scientific domains. Recognizing that entities within these domains can all be represented as sequences, which together form the "language of nature", we introduce Nature Language Model (NatureLM), a sequence-based science foundation model designed for scientific discovery. Pre-trained with data from multiple scientific domains, NatureLM offers a unified, versatile model that enables various applications including: (i) generating and optimizing small molecules, proteins, RNA, and materials using text instructions; (ii) cross-domain generation/design, such as protein-to-molecule and protein-to-RNA generation; and (iii) top performance across different domains, matching or surpassing state-of-the-art specialist models. NatureLM offers a promising generalist approach for various scientific tasks, including drug discovery (hit generation/optimization, ADMET optimization, synthesis), novel material design, and the development of therapeutic proteins or nucleotides. We have developed NatureLM models in different sizes (1 billion, 8 billion, and 46.7 billion parameters) and observed a clear improvement in performance as the model size increases.

cs.AI↗
arXiv2025

On mixed super quasi-Einstein manifolds with Ricci-Bourguignon solitons

This paper delves into the study of mixed super quasi-Einstein manifolds of dimension $n$ (for short, ${\rm M^{n}_{SQE}}$), focusing on their geometric and physical attributes. Initially, we explore several properties of ${\rm M^{n}_{SQE}}$, including conformal Ricci pseudosymmetry, Einstein's field equation, and the space-matter tensor. Subsequently, we characterize mixed super quasi-Einstein manifolds that admit Ricci-Bourguignon solitons. We establish that if the generator vector field is torse-forming, the manifold reduces to a pseudo generalized quasi-Einstein manifold, and we provide a detailed characterization of the eigenvalue problem associated with the symmetric tensor. To substantiate our findings, we construct an example to illustrate the existence of mixed super quasi-Einstein spacetime.

math.DG↗
arXiv2024

XI-DeepONet: An operator learning method for elliptic interface problems

Scientific computing has been an indispensable tool in applied sciences and engineering, where traditional numerical methods are often employed due to their superior accuracy guarantees. However, these methods often encounter challenges when dealing with problems involving complex geometries. Machine learning-based methods, on the other hand, are mesh-free, thus providing a promising alternative. In particular, operator learning methods have been proposed to learn the mapping from the input space to the solution space, enabling rapid inference of solutions to partial differential equations (PDEs) once trained. In this work, we address the parametric elliptic interface problem. Building upon the deep operator network (DeepONet), we propose an extended interface deep operator network (XI-DeepONet). XI-DeepONet exhibits three unique features: (1) The interface geometry is incorporated into the neural network as an additional input, enabling the network to infer solutions for new interface geometries once trained; (2) The level set function associated with the interface geometry is treated as the input, on which the solution mapping is continuous and can be effectively approximated by the deep operator network; (3) The network can be trained without any input-output data pairs, thus completely avoiding the need for meshes of any kind, directly or indirectly. We conduct a comprehensive series of numerical experiments to demonstrate the accuracy and robustness of the proposed method.

math.NA↗
arXiv2023

Breaking Through the Plasma Wavelength Barrier to Extend the Transparency Range of Ultrathin Indium Tin Oxide Films into the Far Infrared

Indium tin oxide (ITO) film, which is the most commonly used transparent conductive film (TCF), has traditionally been believed to be transparent in the visible spectrum but to reflect infrared (IR) light beyond the plasma wavelength ($λ_p$). However, our theoretical analysis challenges this notion by demonstrating that an ultrathin ITO TCF that is thinner than the light's penetration depth, can overcome the transmission barrier at $λ_p$. To validate the theoretical modeling, we have successfully fabricated ITO films that, despite having $λ_p \approx$ 1 $μ$m, remain transparent from 400 nm to 20 $μ$m. This represents the broadest transparency range ever reported for any In$_2$O$_3$-based TCF. The 10-nm-thick ITO TCFs have high visible transmittance (91.0% at 550 nm), low resistivity (5 $\times$ 10$^{-4}$ $Ω\cdot$ cm), and good IR transmittance (averaging 60% over 1.35 $\unicode{x2013}$ 18.35 $μ$m). Their IR transparency facilitates radiative cooling of the underlying circuitry. When an operational resistor is enclosed by commercial ITO TCFs that are 140 nm thick, its temperature increases. However, using 10-nm-thick ITO TCFs instead of the commercial ones can completely avoid this temperature rise. Moreover, attaching a silver grid to a 10-nm-thick ITO TCF can reduce the effective sheet resistance to ~10 $Ω/\square$ at the expense of only ~3% transmittance. This development paves the way for large-scale applications that require low sheet resistance and far-IR transparency.

physics.app-ph↗