arXiv subjects

Lele Wang

Publications and source records attributed to Lele Wang.

At least 19 recordsLinked to original sources

arXiv2026

UnpredictaBench: A Benchmark for Evaluating Distributional Randomness in LLMs

We introduce UnpredictaBench, an evaluation that tests the ability of large language models (LLMs) to capture true underlying distributions. As LLMs are increasingly used as substitutes for other entities (e.g., for humans in economic simulations), the tendency of many models to collapse towards a single plausible answer means a failure to capture the unpredictability of real systems. Recent work on improving output diversity is insufficient for this setting: simulation requires samples that are calibrated to a target distribution, not merely varied outputs. UnpredictaBench isolates a simplified but fundamental version of this problem: sampling outcomes from individual target distributions, including canonical statistical distributions, distributions induced by stochastic programs, and natural-language scenarios that describe random processes. We introduce 448 such problems together with KS@N, a general-purpose evaluation metric that quantifies how well a model outputs approximate black-box target distributions via the Kolmogorov-Smirnov statistical test. This is the rate at which we fail to reject model samples of size N against ground-truth samples, with larger N indicating greater difficulty. Tested across open and proprietary models, we find a large spread in distributional capabilities. For instance, when models generate samples of size 100 (KS@100, our standard metric), scores range from near 0 to over 20%. No model is able to achieve over 40% at KS@100, showing significant headroom in distributional sampling as a capability. Although adding reasoning can somewhat increase scores, we find no immediate solution for this issue. UnpredictaBench shows that even simple distributional simulation remains challenging, making it a necessary first step toward using LLMs as stand-ins for complex systems. Project website and resources are available at https://unpredictabenchmark.github.io/.

cs.CL↗

arXiv2026

VAMPS: Visual-Assisted Mathematical Problem Solving Benchmark

Multimodal large language models are increasingly capable of complex reasoning, yet their performance often degrades when they must externalize a problem through a tool and then reason over the tool's output, specifically when they rely on visual aids. This gap is especially important because real engineering and scientific workflows often rely on visualization tools for analysis, validation, and decision-making. To study this discrepancy, we introduce VAMPS (Visual-Assisted Mathematical Problem Solving), a benchmark for graph-assisted mathematics. VAMPS contains 1,168 multimodal, bilingual multiple-choice question-answer pairs drawn from Iranian University Entrance Exam algebra and calculus problems and expanded with human-reviewed LLM-generated synthetic variants, all selected so that plotting provides a natural solution strategy by revealing intersections, extrema, asymptotes, etc. Designed for both benchmarking and diagnosis, VAMPS goes beyond prior multimodal benchmarks that primarily evaluate reasoning over fixed visual inputs by testing whether a model can benefit from constructing a useful graph and grounding its answer in the resulting visualization. Overall, we found that across a diverse set of models, direct analytical solving surprisingly outperforms tool-enabled visual solving, even on problems where plotting is a natural strategy.

cs.AI↗

arXiv2026

Probabilistic Gradient Coding via Structure-Preserving Sparsification

Gradient coding is a distributed computing technique aiming to provide robustness against slow or non-responsive computing nodes, known as stragglers, while balancing the computational load for responsive computing nodes. Among existing gradient codes, a construction based on combinatorial designs, called BIBD gradient code, achieves the best trade-off between robustness and computational load in the worst-case adversarial straggler setting. However, the range of system parameters for which BIBD gradient codes exist is limited. In this paper, we overcome these limitations by proposing two new probabilistic gradient codes, termed the \emph{Sparse Gaussian} (SG) gradient code and the \emph{Expansion-Preserving} (EP) gradient code. Through probabilistic constructions, the former preserves the combinatorial structure of BIBDs, while the latter preserves key spectral properties. Both codes are based on a common two-step framework: first generating a random matrix and then applying distinct sparsification procedures. The SG gradient code constructs its encoding matrix from a correlated multivariate Gaussian distribution masked by Bernoulli random variables, while the EP gradient code derives its encoding matrix from sparsified expander-like graph structures that preserve key spectral properties. Experimentally, both codes achieve worst-case error performance comparable to that of the BIBD gradient code (when such a code with the same parameters exists). Moreover, they substantially extend the feasible range of system parameters beyond BIBD and soft BIBD gradient codes, offering practical and theoretically grounded solutions for large-scale distributed computing tasks.

cs.IT↗

arXiv2026

When Minor Edits Matter: LLM-Driven Prompt Attack for Medical VLM Robustness in Ultrasound

Ultrasound is widely used in clinical practice due to its portability, cost-effectiveness, safety, and real-time imaging capabilities. However, image acquisition and interpretation remain highly operator dependent, motivating the development of robust AI-assisted analysis methods. Vision-language models (VLMs) have recently demonstrated strong multimodal reasoning capabilities and competitive performance in medical image analysis, including ultrasound. However, emerging evidence highlights significant concerns about their trustworthiness. In particular, adversarial robustness is critical because Med-VLMs operate via natural-language instructions, rendering prompt formulation a realistic and practically exploitable point of vulnerability. Small variations (typos, shorthand, underspecified requests, or ambiguous wording) can meaningfully shift model outputs. We propose a scalable adversarial evaluation framework that leverages a large language model (LLM) to generate clinically plausible adversarial prompt variants via "humanized" rewrites and minimal edits that mimic routine clinical communication. Using ultrasound multiple-choice question answering benchmarks, we systematically assess the vulnerability of SOTA Med-VLMs to these attacks, examine how attacker LLM capacity influences attack success, analyze the relationship between attack success and model confidence, and identify consistent failure patterns across models. Our results highlight realistic robustness gaps that must be addressed for safe clinical translation. Code will be released publicly following the review process.

cs.CV↗

arXiv2026

Toward Agentic AI: Task-Oriented Communication for Hierarchical Planning of Long-Horizon Tasks

Agentic artificial intelligence (AI) is an AI paradigm that can perceive the environment, reason over observations, and execute actions to achieve specific goals. Task-oriented communication supports agentic AI by transmitting only the task-related information instead of full raw data in order to reduce the bandwidth requirement. In real-world scenarios, AI agents often need to perform a sequence of actions to complete complex tasks. Completing these long-horizon tasks requires a hierarchical agentic AI architecture, where a high-level planner module decomposes a task into subtasks, and a low-level actor module executes each subtask sequentially. Since each subtask has a distinct goal, the existing task-oriented communication schemes are not designed to handle different goals for different subtasks. To address this challenge, in this paper, we develop a hierarchical task-oriented communication (HiTOC) framework. We consider a system with an edge server and a robot as an edge device. The high-level planner and low-level actor modules reside on the edge server. The robot transmits only the environmental observation that is relevant to the current subtask to the edge server. We propose a conditional variational information bottleneck (cVIB) approach to train the HiTOC framework to adaptively transmit minimal information required for each subtask. Simulations conducted on the AI2-THOR platform demonstrate that the proposed HiTOC framework outperforms three state-of-the-art schemes in terms of the success rate on MAP-THOR benchmark.

eess.IV↗

arXiv2026

On the Generalization Error of Differentially Private Algorithms via Typicality

We study the generalization error of stochastic learning algorithms from an information-theoretic perspective, with a particular emphasis on deriving sharper bounds for differentially private algorithms. It is well known that the generalization error of stochastic learning algorithms can be bounded in terms of mutual information and maximal leakage, yielding in-expectation and high-probability guarantees, respectively. In this work, we further upper bound mutual information and maximal leakage by explicit, easily computable formulas, using typicality-based arguments and exploiting the stability properties of private algorithms. In the first part of the paper, we strictly improve the mutual-information bounds by Rodr\'iguez-G\'alvez et al. (IEEE Trans. Inf. Theory, 2021). In the second part, we derive new upper bounds on the maximal leakage of learning algorithms. In both cases, the resulting bounds on information measures translate directly into generalization error guarantees.

cs.IT↗

arXiv2026

Information-Theoretic Limits on Exact Subgraph Alignment Problem

The graph alignment problem aims to identify the vertex correspondence between two correlated graphs. Most existing studies focus on the scenario in which the two graphs share the same vertex set. However, in many real-world applications, such as computer vision, social network analysis, and bioinformatics, the task often involves locating a small graph pattern within a larger graph. Existing graph alignment algorithms and analysis cannot directly address these scenarios because they are not designed to identify the specific subset of vertices where the small graph pattern resides within the larger graph. Motivated by this limitation, we introduce the subgraph alignment problem, which seeks to recover both the vertex set and/or the vertex correspondence of a small graph pattern embedded in a larger graph. In the special case where the small graph pattern is an induced subgraph of the larger graph and both the vertex set and correspondence are to be recovered, the problem reduces to the subgraph isomorphism problem, which is NP-complete in the worst case. In this paper, we formally formulate the subgraph alignment problem by proposing the Erdos-Renyi subgraph pair model together with some appropriate recovery criterion. We then establish almost-tight information-theoretic results for the subgraph alignment problem and present some novel approaches for the analysis.

cs.IT↗

arXiv2025

Test-Time Steering for Lossless Text Compression via Weighted Product of Experts

Lossless compression techniques are crucial in an era of rapidly growing data. Traditional universal compressors like gzip offer low computational overhead, high speed, and broad applicability across data distributions. However, they often lead to worse compression rates than modern neural compressors, which leverage large-scale training data to model data distributions more effectively. Despite their advantages, neural compressors struggle to generalize to unseen data. To address this limitation, we propose a novel framework that performs Test-Time Steering via a Weighted Product of Experts (wPoE). At inference, our method adaptively combines a universal compression model with a pretrained neural language model, ensuring the compression rate is at least as good as that of the best individual model. Extensive experiments demonstrate that our approach improves the performance of text compression without requiring fine-tuning. Furthermore, it seamlessly integrates with any autoregressive language model, providing a practical solution for enhancing text compression across diverse data distributions.

cs.CL↗

arXiv2025

StreamSplat: Towards Online Dynamic 3D Reconstruction from Uncalibrated Video Streams

Real-time reconstruction of dynamic 3D scenes from uncalibrated video streams demands robust online methods that recover scene dynamics from sparse observations under strict latency and memory constraints. Yet most dynamic reconstruction methods rely on hours of per-scene optimization under full-sequence access, limiting practical deployment. In this work, we introduce StreamSplat, a fully feed-forward framework that instantly transforms uncalibrated video streams of arbitrary length into dynamic 3D Gaussian Splatting (3DGS) representations in an online manner. It is achieved via three key technical innovations: 1) a probabilistic sampling mechanism that robustly predicts 3D Gaussians from uncalibrated inputs; 2) a bidirectional deformation field that yields reliable associations across frames and mitigates long-term error accumulation; 3) an adaptive Gaussian fusion operation that propagates persistent Gaussians while handling emerging and vanishing ones. Extensive experiments on standard dynamic and static benchmarks demonstrate that StreamSplat achieves state-of-the-art reconstruction quality and dynamic scene modeling. Uniquely, our method supports the online reconstruction of arbitrarily long video streams with a 1200x speedup over optimization-based methods. Our code and models are available at https://streamsplat3d.github.io/.

cs.CV↗

arXiv2025

On the Worst-Case Complexity of Gibbs Decoding for Reed--Muller Codes

Reed--Muller (RM) codes are known to achieve capacity on binary symmetric channels (BSC) under the Maximum a Posteriori (MAP) decoder. However, it remains an open problem to design a capacity achieving polynomial-time RM decoder. Due to a lemma by Liu, Cuff, and Verd\'u, it can be shown that decoding by sampling from the posterior distribution is also capacity-achieving for RM codes over BSC. The Gibbs decoder is one such Markov Chain Monte Carlo (MCMC) based method, which samples from the posterior distribution by flipping message bits according to the posterior, and can be modified to give other MCMC decoding methods. In this paper, we analyze the mixing time of the Gibbs decoder for RM codes. Our analysis reveals that the Gibbs decoder can exhibit slow mixing for certain carefully constructed sequences. This slow mixing implies that, in the worst-case scenario, the decoder requires super-polynomial time to converge to the desired posterior distribution.

cs.IT↗

arXiv2025

MoFlow: One-Step Flow Matching for Human Trajectory Forecasting via Implicit Maximum Likelihood Estimation based Distillation

In this paper, we address the problem of human trajectory forecasting, which aims to predict the inherently multi-modal future movements of humans based on their past trajectories and other contextual cues. We propose a novel motion prediction conditional flow matching model, termed MoFlow, to predict K-shot future trajectories for all agents in a given scene. We design a novel flow matching loss function that not only ensures at least one of the $K$ sets of future trajectories is accurate but also encourages all $K$ sets of future trajectories to be diverse and plausible. Furthermore, by leveraging the implicit maximum likelihood estimation (IMLE), we propose a novel distillation method for flow models that only requires samples from the teacher model. Extensive experiments on the real-world datasets, including SportVU NBA games, ETH-UCY, and SDD, demonstrate that both our teacher flow model and the IMLE-distilled student model achieve state-of-the-art performance. These models can generate diverse trajectories that are physically and socially plausible. Moreover, our one-step student model is $\textbf{100}$ times faster than the teacher flow model during sampling. The code, model, and data are available at our project page: https://moflow-imle.github.io

cs.CV↗

arXiv2025

SimMark: A Robust Sentence-Level Similarity-Based Watermarking Algorithm for Large Language Models

The widespread adoption of large language models (LLMs) necessitates reliable methods to detect LLM-generated text. We introduce SimMark, a robust sentence-level watermarking algorithm that makes LLMs' outputs traceable without requiring access to model internals, making it compatible with both open and API-based LLMs. By leveraging the similarity of semantic sentence embeddings combined with rejection sampling to embed detectable statistical patterns imperceptible to humans, and employing a soft counting mechanism, SimMark achieves robustness against paraphrasing attacks. Experimental results demonstrate that SimMark sets a new benchmark for robust watermarking of LLM-generated content, surpassing prior sentence-level watermarking techniques in robustness, sampling efficiency, and applicability across diverse domains, all while maintaining the text quality and fluency.

cs.CL↗

arXiv2024

Low-Complexity Coding Techniques for Cloud Radio Access Networks

The problem of coding for the uplink and downlink of cloud radio access networks (C-RAN's) with $K$ users and $L$ relays is considered. It is shown that low-complexity coding schemes that achieve any point in the rate-fronthaul region of joint coding and compression can be constructed starting from at most $4(K+L)-2$ point-to-point codes designed for symmetric channels. This reduces the seemingly hard task of constructing good codes for C-RAN's to the much better understood task of finding good codes for single-user channels. To show this result, an equivalence between the achievable rate-fronthaul regions of joint coding and successive coding is established. Then, rate-splitting and quantization-splitting techniques are used to show that the task of achieving a rate-fronthaul point in the joint coding region can be simplified to that of achieving a corner point in a higher-dimensional C-RAN problem. As a by-product, some interesting properties of the rate-fronthaul region of joint decoding for uplink C-RAN's are also derived.

cs.IT↗

arXiv2024

Fr\'echet Video Motion Distance: A Metric for Evaluating Motion Consistency in Videos

Significant advancements have been made in video generative models recently. Unlike image generation, video generation presents greater challenges, requiring not only generating high-quality frames but also ensuring temporal consistency across these frames. Despite the impressive progress, research on metrics for evaluating the quality of generated videos, especially concerning temporal and motion consistency, remains underexplored. To bridge this research gap, we propose Fr\'echet Video Motion Distance (FVMD) metric, which focuses on evaluating motion consistency in video generation. Specifically, we design explicit motion features based on key point tracking, and then measure the similarity between these features via the Fr\'echet distance. We conduct sensitivity analysis by injecting noise into real videos to verify the effectiveness of FVMD. Further, we carry out a large-scale human study, demonstrating that our metric effectively detects temporal noise and aligns better with human perceptions of generated video quality than existing metrics. Additionally, our motion features can consistently improve the performance of Video Quality Assessment (VQA) models, indicating that our approach is also applicable to unary video quality evaluation. Code is available at https://github.com/ljh0v0/FMD-frechet-motion-distance.

cs.CV↗

arXiv2024

Optical skyrmions from metafibers

Optical skyrmions are an emerging class of structured light with sophisticated particle-like topologies with great potential for revolutionizing modern informatics. However, the current generation of optical skyrmions involves complex or bulky systems, hindering their development of practical applications. Here, exploiting the emergent "lab-on-fiber" technology, we demonstrate the design of a metafiber-integrated photonic skyrmion generator. We not only successfully generated high-quality optical skyrmions from metafibers, but also experimentally verified their remarkable properties, such as regulability and topological stability with deep-subwavelength features beyond the diffraction limits. Our flexible and fiber-integrated optical skyrmions platform paves the avenue for future applications of topologically-enhanced remote super-resolution microscopy and super-robust information transfer.

physics.optics↗

arXiv2024

BCAmirs at SemEval-2024 Task 4: Beyond Words: A Multimodal and Multilingual Exploration of Persuasion in Memes

Memes, combining text and images, frequently use metaphors to convey persuasive messages, shaping public opinion. Motivated by this, our team engaged in SemEval-2024 Task 4, a hierarchical multi-label classification task designed to identify rhetorical and psychological persuasion techniques embedded within memes. To tackle this problem, we introduced a caption generation step to assess the modality gap and the impact of additional semantic information from images, which improved our result. Our best model utilizes GPT-4 generated captions alongside meme text to fine-tune RoBERTa as the text encoder and CLIP as the image encoder. It outperforms the baseline by a large margin in all 12 subtasks. In particular, it ranked in top-3 across all languages in Subtask 2a, and top-4 in Subtask 2b, demonstrating quantitatively strong performance. The improvement achieved by the introduced intermediate step is likely attributable to the metaphorical essence of images that challenges visual encoders. This highlights the potential for improving abstract visual semantics encoding.

cs.CL↗

arXiv2024

An Information-Theoretic Framework for Out-of-Distribution Generalization with Applications to Stochastic Gradient Langevin Dynamics

We study the Out-of-Distribution (OOD) generalization in machine learning and propose a general framework that establishes information-theoretic generalization bounds. Our framework interpolates freely between Integral Probability Metric (IPM) and $f$-divergence, which naturally recovers some known results (including Wasserstein- and KL-bounds), as well as yields new generalization bounds. Additionally, we show that our framework admits an optimal transport interpretation. When evaluated in two concrete examples, the proposed bounds either strictly improve upon existing bounds in some cases or match the best existing OOD generalization bounds. Moreover, by focusing on $f$-divergence and combining it with the Conditional Mutual Information (CMI) methods, we derive a family of CMI-based generalization bounds, which include the state-of-the-art ICIMI bound as a special instance. Finally, leveraging these findings, we analyze the generalization of the Stochastic Gradient Langevin Dynamics (SGLD) algorithm, showing that our derived generalization bounds outperform existing information-theoretic generalization bounds in certain scenarios.

cs.IT↗

arXiv2024

Noisy Computing of the Threshold Function

Let $\mathsf{TH}_k$ denote the $k$-out-of-$n$ threshold function: given $n$ input Boolean variables, the output is $1$ if and only if at least $k$ of the inputs are $1$. We consider the problem of computing the $\mathsf{TH}_k$ function using noisy readings of the Boolean variables, where each reading is incorrect with some fixed and known probability $p \in (0,1/2)$. As our main result, we show that it is sufficient to use $(1+o(1)) \frac{n\log \frac{m}{\delta}}{D_{\mathsf{KL}}(p \| 1-p)}$ queries in expectation to compute the $\mathsf{TH}_k$ function with a vanishing error probability $\delta = o(1)$, where $m\triangleq \min\{k,n-k+1\}$ and $D_{\mathsf{KL}}(p \| 1-p)$ denotes the Kullback-Leibler divergence between $\mathsf{Bern}(p)$ and $\mathsf{Bern}(1-p)$ distributions. Conversely, we show that any algorithm achieving an error probability of $\delta = o(1)$ necessitates at least $(1-o(1))\frac{(n-m)\log\frac{m}{\delta}}{D_{\mathsf{KL}}(p \| 1-p)}$ queries in expectation. The upper and lower bounds are tight when $m=o(n)$, and are within a multiplicative factor of $\frac{n}{n-m}$ when $m=\Theta(n)$. In particular, when $k=n/2$, the $\mathsf{TH}_k$ function corresponds to the $\mathsf{MAJORITY}$ function, in which case the upper and lower bounds are tight up to a multiplicative factor of two. Compared to previous work, our result tightens the dependence on $p$ in both the upper and lower bounds.

cs.DS↗