Wei Li

The paper presents CAPE, a framework that generates natural-language explanations for spatially organized document layouts, using context-aware representations and LLM-based explanation generation.

This paper analyzes Bregman ADMM for nonconvex linearly constrained problems under two-sided relative smoothness, showing convergence to strict saddle points and almost-sure second-order stationarity.

The paper proposes a framework to harvest unused computation resources on AI chips for general-purpose tasks using neural architecture search and approximation techniques.

This paper presents Clarus, a collaboration infrastructure for coordinating autonomous research agents towards web-scale scientific collaboration.

This paper studies the distribution of reinforcement learning (RL) adaptation across transformer layers in large language models and finds that training a single layer can recover most of the gains obtained during full RL training.

The paper presents SenseNova-Vision, a unified multimodal model for computer vision tasks using natural language instructions and optional visual prompts, trained primarily on a new corpus and requiring no task-specific modifications.

The paper introduces SLORR, a simple and stateless framework for in-training low-rank regularization of neural networks using two variants based on the Hoyer sparsity metric and the nuclear norm.

This paper introduces ZipLine, a system for integrative analysis of multivariate graphs through a unified predicate language and learning algorithm.

The paper presents teLLMe, a system for exploratory causal analysis of urban driving datasets using structured event tables, causal structure learning, and query-specific effect estimation.

xDSM is a full-space, elastic DSM system built over CXL that transparently scales unmodified multithreaded applications by employing an OS-runtime co-design, dynamic data placement policy, and spatial locality-aware elasticity.

This paper proposes a Kalman filter-assisted data-predictive SAR ADC to reduce switching energy and latency in ultra-low-power IoT devices.

This paper presents an efficient fault-tolerance scheme for CPU-based fully homomorphic encryption (FHE) that reduces protection overhead and achieves 100 percent detection rate under random single-bit transient faults.

This paper introduces Isospectral Optimization (ISO), a new optimization framework for reinforcement learning with verifiable rewards (RLVR), which inherits the base model's weight spectra while optimizing the frames to achieve stronger performance.

A new framework, PRTA, is proposed for full-ranking recommendation tasks using large language models, where an LLM acts as a central planner and traditional recommendation models perform scoring.

This paper proposes CoHarden, a co-generation framework for automated program repair that uses a lax signal as an in-loop convergence criterion to prevent lax regressions.

This paper proposes multi-carrier positioning frameworks for pinching-antenna systems using matrix pencil-based and Rank-1 ranging algorithms, and a two-stage weighted nonlinear least-squares positioning algorithm.

This paper constructs strictly convex quadratic problems and initial points for which the long Barzilai--Borwein method does not converge root-superlinearly.

This paper presents HarnessLLM, an automated workflow that uses large language models to generate verification harnesses for Rust code from test suites, improving memory safety.

The paper introduces SceneActBench, a benchmark for evaluating vision-language model agents' ability to perform actions on multi-object 3D scenes.

This paper organizes embodied data sources for multimodal foundation models into a pyramid, focusing on real-robot, UMI-style, egocentric and exocentric, simulation, and general vision-language data.