Zero-real Agent Policy (ZAP), a recipe that teaches a small off-the-shelf vision-language model to control a robot via a simple tool harness, uses a pretrained 27B parameter model to bootstrap world understanding and reasoning, followed by supervised fine-tuning and GRPO on simple simulated tasks to teach it how to mov...
Benjamin Pikus, D. Saunders, Mircea Mironenco et al.· 0 citations
Multi-fidelity training, which combines abundant low-fidelity data with scarce high-fidelity labels, is increasingly common in atomistic modeling, but its effect on predictive uncertainty remains largely unexplored. We investigate this question using deep ensembles of mean-variance MACE models trained on ANI-1x, contra...
Matteo Tolloso, Davide Bacciu, Luigi Bonati· 0 citations
Web services increasingly expose backend interfaces that let users delegate tasks to AI agents, which invoke service operations directly without navigating websites. This shift changes how existing Web vulnerabilities can be exploited. We investigate these changes through VELA, an executable benchmark covering 44 vulne...
Ruozhao Yang, Jiongchi Yu, Ming Hu et al.· 0 citations
Unsafe deserialization turns untrusted data into attacker-controlled execution, and a single vulnerability can admit diverse exploit variants that evade defenses tailored to a known exploit. Existing defenses typically operate at a single enforcement point with limited visibility into the full attack. Combining control...
Ruo-Zhao Yang, Ming-Lang Li, Jiongchi Yu et al.· 0 citations
Reach audiences
Advertise in front of researchers, engineers, and readers.
High-performance General Matrix Multiplication (GEMM) implementations rely on breaking down GEMM into carefully selected tiles to achieve peak performance. We present Waypoint, a new analytical model to automatically identify the optimal tile size for a given hardware and input matrix shape. Waypoint estimates the achi...
N. Srinidhi, Sunanda Somwase, Hemant Kumar Chandrakant Chandavar et al.· 0 citations
CACHELINT is designed, which models a prompt-caching mis-integration as an invocation site together with the preconditions required for correct cache integration, and locates caching invocation sites with CodeQL and discharges each precondition either against a dated provider capability oracle or with an LLM guard chec...
Ying Zhang, Ruo-Qi Guo, Yi Liu et al.· 0 citations
VLA-HWBench is introduced, a benchmark that measures 18 open VLA models on NVIDIA Jetson AGX Thor and AMD Ryzen AI Max+ PRO 395, with a seven-stage inference breakdown, and reveals three insights that suggest low-latency edge VLAs should use few action-generation iterations, inference software should extend kernel fusi...
Bo-Rui Xu, Hong-Shi Tan, Wen-Qi Jiang et al.· 1 citation
Science agents increasingly package domain expertise as reusable skills that plan analyses, select tools, and execute workflows. This modularity creates a confidentiality risk: the interactions that deliver a skill's capability can also reveal the skill itself. We build an extensible benchmark of 40 protected skills ac...
Cheng Gui, Xiang Zheng, Yutao Wu et al.· 0 citations
Visual grounding (VG) in visual question answering (VQA) is a fundamental capability of Vision-Language Models (VLMs), enabling the detection of the most relevant region in an image to answer the question. This reflects the model's capability of multimodal comprehension, which is crucial when interacting with visually...
Fairooz Azim, Lydia Achour, Claudia Borg et al.· 0 citations
On-device fine-tuning lets a large language model (LLM) learn from its user's private data without uploading it. Low-rank adaptation (LoRA) reduces the trainable parameters, but each first-order training step still multiplies by the frozen base-model weights during the forward pass, activation recomputation, and input-...
Xu Zheng, Wei Cheng, Zhuo-Min Chen et al.· 1 citation
Full-field reconstruction of air pollution is essential for evaluating pollution exposure and supporting public health decision-making. However, the complex interactions among pollutants, hard-to-predict weather patterns, and limited monitoring station coverage make this a complex task. We apply deep learning technique...
Abhishek A. Sabnis, Mihai Mitrea, Lya Lugon et al.· 0 citations
This work presents Trust Region Policy Distillation (TOP-D), which transforms the notoriously unstable, high-variance On-Policy Distillation (OPD) into a stable training paradigm by dynamically constructing a proximal teacher and establishes a rigorous framework demonstrating that TOP-D inherently controls gradient var...
Zheng-Peng Xie, L. Zhang, Ze-Ke Xie et al.· arXiv.org· 0 citations
