Virtual Technology (United States)

Text-attributed graphs (TAGs) couple graph topology with node-level text, but real data often contain spurious edges, missing links, and text–structure mismatch that destabilize learning under scarce labels. We propose CoACL (Coupled Augmentation for Contrastive Learning), a framework that uses LLM semantic supervision to denoise structural and textual information and alleviate data sparsity. CoACL first prunes the candidate edge space using structural similarity and then queries an LLM to discard suspicious edges and confirm plausible links, yielding semantically consistent positive and negative pairs. We further introduce keyword-focused text augmentations and learn coupled representations by optimizing a joint text–graph contrastive objective guided by semantics. Experiment results on Cora, PubMed, and the Open Graph Benchmark Arxiv dataset (OGBN-Arxiv) show that CoACL consistently outperforms strong baselines and yields up to 7.1% absolute improvement in node classification accuracy, with the largest gains in low-label regimes. By constraining LLM evaluation to similarity-based candidates, CoACL targets neighborhood-level noise with controlled cost.

An R&D initiative within Aardman Animations exploring VP technologies for stop-motion film making. Taking a holistic view of VP from story development through to delivery. Utilising real-time tools, digital twins, and a cross platform XR sandbox. Striving to evolve traditional processes, enhancing creativity, efficiency, and integration across the production pipeline.

Flexible pressure sensors are in the spotlight for various fields, such as human–machine interfaces and intelligent robotics. In this study, we designed a gradient pore density (GPD) based polypyrrole (Ppy)/polydimethylsiloxane (PDMS) sponge for a flexible pressure sensor by using a fused deposition modeling type three-dimensional printer. The GPD PDMS sponge was fabricated by using an acrylonitrile butadiene styrene filament mold with gradient infill density. PDMS was then cast and cured, and the mold was dissolved in acetone. In situ polymerization was used to coat the Ppy onto the GPD PDMS sponge and produced the GPD Ppy/PDMS sponge pressure sensor (GPPS). GPPS showed a wide sensing range of 0–1500 kPa, a maximum sensitivity of 2.26 kPa–1, and a low limit of detection of 17 Pa. It also proved excellent sensing stability across loading rates of 0.1–30 mm/min, and cyclic durability after 2000 repeated compression-recovery tests at two different pressure levels. Additionally, GPPS was confirmed to be applicable as a wearable sensor across a range from low to high pressure. The spatial resolution of the sensor was demonstrated through various tests using a 3 × 3 sensor array, which confirms its applicability in multi-information sensing fields.

The rapid advancements in autonomous driving technology necessitate the extensive deployment of automotive radars operating within the 77-81 GHz millimeter-wave band in the forthcoming years. In contrast to earlier Frequency Modulated Continuous Wave (FMCW) radars, Phase-Modulated Continuous Wave (PMCW) radars exhibit notable improvements in processing speed and flexibility, offering superior range and velocity resolution. These enhancements are critical for the precise detection and interpretation of dynamic and complex traffic scenarios. This paper initially presents the simulation and testing of algorithms for range and speed measurement using single-input single-output (SISO) PMCW millimeterwave radar. Building upon these results, further simulations incorporating multi-input multi-output (MIMO) systems utilizing Hadamard codes are conducted to augment PMCW radar performance. The final phase of this study involves implementing the relevant signal processing algorithms on a custom-developed digital signal processor (DSP) named SWIFT. Experimental findings demonstrate that PMCW radar significantly mitigates multipath effects and clutter, maintaining high performance in complex environments. Furthermore, the algorithms executed on the DSP meet the anticipated performance standards. The proposed methodology not only validates the theoretical framework but also establishes a foundation for future hardware implementation.

Large Language Models (LLMs) excel in many Natural Language Processing tasks but are outperformed by specialized tools for certain tasks.This raises the question: Can we reduce redundant LLM parameters when using these tools?Given the size and high training costs of LLMs, it is essential to efficiently forget specific knowledge without retraining.This paper introduces UNLEARN, a novel method that uses subspace techniques to selectively remove knowledge without access to the original training data, without retraining, and with minimal impact to other tasks.Our results show that UNLEARN significantly outperforms previous methods for forgetting targeted (unwanted) knowledge while also preserving related (wanted) knowledge.We also propose LEARN, a complementary approach for targeted knowledge addition, which achieves fine-tuning accuracy comparable to Low-Rank Adaptation (LoRA) without degrading related task performance. 1