Graph contrastive learning (GCL) has demonstrated great promise for learning generalizable graph representations from unlabeled data. However, conventional GCL approaches face two critical limitations: (1) the restricted expressive capacity of multilayer perceptron (MLP) based encoders, and (2) suboptimal negative samples that either from random augmentations—failing to provide effective 'hard negatives'—or generated hard negatives without addressing the semantic distinctions crucial for discriminating graph data. To this end, we propose Khan-GCL, a novel framework that integrates the Kolmogorov–Arnold Network (KAN) into the GCL encoder architecture, substantially enhancing its representational capacity. Furthermore, we exploit the rich information embedded within KAN coefficient parameters to develop two novel critical feature identification techniques that enable the generation of semantically meaningful hard negative samples for each graph representation. These strategically constructed hard negatives guide the encoder to learn more discriminative features by emphasizing critical semantic differences between graphs. Extensive experiments demonstrate that our approach achieves state-of-the-art performance compared to existing GCL methods across a variety of datasets and tasks.

Visual autoregressive modeling (VAR) via next-scale prediction has emerged as a scalable image generation paradigm. While Key and Value (KV) caching in large language models (LLMs) has been extensively studied, next-scale prediction presents unique challenges, and KV caching design for next-scale based VAR transformers remains largely unexplored. A major bottleneck is the excessive KV memory growth with the increasing number of scales—severely limiting scalability. Our systematic investigation reveals that: (1) Attending to tokens from local scales significantly contributes to generation quality (2) Allocating a small amount of memory for the coarsest scales, termed as condensed scales, stabilizes multi-scale image generation (3) Strong KV similarity across finer scales is predominantly observed in cache-efficient layers, whereas cache-demanding layers exhibit weaker inter-scale similarity. Based on the observations, we introduce AMS-KV, a scale-adaptive KV caching policy for next-scale prediction in VAR models. AMS-KV prioritizes storing KVs from condensed and local scales, preserving the most relevant tokens to maintain generation quality. It further optimizes KV cache utilization and computational efficiency identifying cache-demanding layers through inter-scale similarity analysis. Compared to the vanilla next-scale prediction-based VAR models, AMS-KV reduces KV cache usage by up to 84.83% and self-attention latency by 60.48%. Moreover, when the baseline VAR-d30 model encounters out-of-memory failures at a batch size of 128, AMS-KV enables stable scaling to a batch size of 256 with improved throughput.

This paper introduces Learn-by-Compare (LbC), a novel approach for analog performance modeling by employing semi-supervised contrastive regression. LbC employs a deep neural network encoder to come up with latent representations of sizing solutions by comparing similarity/dissimilarity of the underlying performance. Leveraging two levels of transistor level sizing data augmentation (DA), namely LS-DA and GS-DA, LbC produces new data samples by employing design knowledge. Experimental results highlight LbC's superior predictive accuracy compared to traditional regression methods. Offering a streamlined semi-supervised learning methodology, LbC effectively incorporates simple design knowledge and representation learning for efficient analog performance modeling.

Wafer map analysis is essential for process issue detection and yield improvement in semiconductor manufacturing. Accurate wafer map pattern recognition facilitates root-causing of abnormal chip fabrication conditions. However, manually annotating wafer map data is expensive and time-consuming, which drives up the demand for exploring label-efficient methods for wafer analysis. This paper proposes a novel contrastive learning framework for wafer map pattern feature extraction and classification. Under the semi-supervised learning setting, the proposed approach aims at learning from a large amount of unlabeled data while efficiently exploiting a small amount of expensive labeled data. Our method utilizes supervised contrastive learning on a small amount of labeled data to learn a better latent space representation with well-separated wafer pattern classes. Furthermore, a dual-encoder latent-space model is incorporated to best optimize the simultaneous use of labeled, unlabeled data, and varying types of data augmentations for representation learning. Finally, we enrich the semantics of the learned latent representation space by introducing a novel interwafer data augmentation to synthesize data which are not present in the given dataset. Experiments show that our method leads existing wafer pattern recognition techniques including recent contrastive learning based approaches by a large performance gain, and suggest that superior accuracy may be achieved simply by semi-supervised learning without resorting to labeling-intensive supervised learning.