We present an efficient algorithm for regularized optimal transport.

Knowledge-based conversational question answering (KBCQA) confronts persistent challenges in resolving coreference, modeling contextual dependencies, and executing complex logical reasoning. Existing approaches, whether end-to-end semantic parsing or stepwise agent-based reasoning--often suffer from structural inaccuracies and prohibitive computational costs, particularly when processing intricate queries over large knowledge graphs. To address these limitations, we introduce SEAL, a novel two-stage semantic parsing framework grounded in self-evolving agentic learning. This core is then refined by an agentic calibration module, which corrects syntactic inconsistencies and aligns entities and relations precisely with the underlying knowledge graph. This decomposition not only simplifies logical form generation but also significantly enhances structural fidelity and linking efficiency. Crucially, SEAL incorporates a self-evolving mechanism that integrates local and global memory with a reflection module, enabling continuous adaptation from dialog history and execution feedback without explicit retraining. Extensive experiments on the SPICE benchmark demonstrate that SEAL achieves state-of-the-art performance, especially in multi-hop reasoning, comparison, and aggregation tasks. Introduction A Knowledge Graph (KG) is a structured representation of knowledge, typically organized as triples (head entity, relation, tail entity) to encode factual information [1]. In recent years, KGs have gained widespread attention in both academia and industry [2, 3]. Knowledge-based Question Answering (KBQA) systems are designed to query these structured KGs, using reasoning to provide accurate answers to natural language questions [4, 5]. Among KBQA methods, Semantic Parsing (SP) based approaches translate questions into structured queries (e.g., SPARQL, Cypher, etc.) for execution against the KG, offering strong interpretability and high efficiency [6, 7]. These systems are widely applied in fields such as healthcare and business, significantly reducing the technical threshold for accessing complex knowledge systems. Knowledge-based conversational QA (KBCQA) extends this paradigm to multi-turn interactive scenarios, requiring the system to conduct continuous reasoning and to address dialog understanding challenges such as coreference resolution [8, 9]. For this task, SP remains a mainstream approach, where the goal is to convert contextual natural language queries into executable logical forms. While LLMs offer significant opportunities for SP-based KBQA, and KBCQA tasks, current methods face substantial limitations in handling struc-2 turally complex questions [15].

The Kolmogorov-Arnold Network (KAN) has been gaining popularity as an alternative to the multi-layer perceptron (MLP) with its increased expressiveness and interpretability. Even so, the KAN suffers from being orders of magnitude slower due to its increased computational cost and training instability, limiting its applicability to larger-scale tasks. Recently, the Kolmogorov-Arnold Transformer (KAT) has been proposed, which can achieve FLOPs similar to the traditional Transformer with MLPs by leveraging Group-Rational KAN (GR-KAN). Unfortunately, despite the comparable FLOPs, our testing reveals that the KAT is still 123x slower in training speeds, indicating that there are other performance bottlenecks beyond FLOPs. In this paper, we conduct a series of experiments to understand the root cause of the slowdown in KAT. We uncover that the slowdown can be isolated to memory stalls, linked more specifically to inefficient gradient accumulations in the backward pass of GR-KAN. To address this memory bottleneck, we propose FlashKAT, which minimizes accesses to slow memory and the usage of atomic adds through a restructured kernel. Evaluations demonstrate that FlashKAT can achieve a training speedup of 86.5x compared with the state-of-the-art KAT, while reducing rounding errors in the computation of the gradients.

Recent leaps in large language models (LLMs) caused a revolution in programming tools (like GitHub Copilot) that can help with code generation, debugging, and even performance optimization. In this paper, we focus on the capabilities of the most recent reasoning models to generate optimized CUDA code for predefined, well-known tasks. Our objective is to determine which types of code optimizations and parallel patterns the LLMs can perform by themselves and whether they can be improved by tutoring (providing more detailed hints and guidelines in the prompt). The generated solutions were evaluated both automatically (for correctness and speedup) and manually (code reviews) to provide a more detailed perspective. We also tried an interactive approach where the LLM can fix its previous mistakes within a session. The results indicate that LLMs are quite skilled coders; however, they require tutoring to reach optimized solutions provided by parallel computing experts.

We present an efficient algorithm for regularized optimal transport.

Large language model (LLM) personalization aims to tailor model behavior to individual users based on their historical interactions. However, its effectiveness is often hindered by two key challenges: the \textit{cold-start problem}, where users with limited history provide insufficient context for accurate personalization, and the \textit{biasing problem}, where users with abundant but skewed history cause the model to overfit to narrow preferences. We identify both issues as symptoms of a common underlying limitation, i.e., the inability to model collective knowledge across users. To address this, we propose a local-global memory framework (LoGo) that combines the personalized local memory with a collective global memory that captures shared interests across the population. To reconcile discrepancies between these two memory sources, we introduce a mediator module designed to resolve conflicts between local and global signals. Extensive experiments on multiple benchmarks demonstrate that LoGo consistently improves personalization quality by both warming up cold-start users and mitigating biased predictions. These results highlight the importance of incorporating collective knowledge to enhance LLM personalization.

Serving Large Language Models (LLMs) is critical for AI-powered applications, yet it demands substantial computational resources, particularly in memory bandwidth and computational throughput. Low-precision computation has emerged as a key technique to improve efficiency while reducing resource consumption. Existing approaches for generating low-precision kernels are limited to weight bit widths that are powers of two and suffer from suboptimal performance because of high-level GPU programming abstractions. These abstractions restrict critical optimizations, such as fine-grained register management and optimized memory access patterns, that are essential for efficient low-precision computations. In this paper, we introduce Tilus, a domain-specific language designed for General-Purpose GPU (GPGPU) computing that supports low-precision data types with arbitrary bit widths from 1 to 8 while maintaining GPU programmability. Tilus features a thread-block-level programming model, a hierarchical memory space, a novel algebraic layout system, and extensive support for diverse low-precision data types. Tilus programs are compiled into highly efficient GPU programs through automatic vectorization and instruction selection. Extensive experiments demonstrate that Tilus efficiently supports a full spectrum of low-precision data types, and outperforms state-of-the-art low-precision kernels. Compared to existing compilers such as Triton and Ladder, as well as hand-optimized kernels such as QuantLLM and Marlin, Tilus achieves performance improvements of: $1.75\times$, $2.61\times$, $1.29\times$ and $1.03\times$, respectively. We open-source Tilus at https://github.com/NVIDIA/tilus.

Abstract--This paper presents J3DAI, a tiny deep neural network-based hardware accelerator for a 3-layer 3D-stacked CMOS image sensor featuring an artificial intelligence (AI) chip integrating a Deep Neural Network (DNN)-based accelerator . The DNN accelerator is designed to efficiently perform neural network tasks such as image classification and segmentation. This paper focuses on the digital system of J3DAI, highlighting its Performance-Power-Area (PPA) characteristics and showcasing advanced edge AI capabilities on a CMOS image sensor . T o support hardware, we utilized the Aidge comprehensive software framework, which enables the programming of both the host processor and the DNN accelerator . Aidge supports post-training quantization, significantly reducing memory footprint and computational complexity, making it crucial for deploying models on resource-constrained hardware like J3DAI. Our experimental results demonstrate the versatility and efficiency of this innovative design in the field of edge AI, showcasing its potential to handle both simple and computationally intensive tasks. Future work will focus on further optimizing the architecture and exploring new applications to fully leverage the capabilities of J3DAI. As edge AI continues to grow in importance, innovations like J3DAI will play a crucial role in enabling real-time, low-latency, and energy-efficient AI processing at the edge. The increasing adoption of intelligent vision systems in various domains, including the Internet of Things (IoT), healthcare, automotive applications, and smart surveillance, has led to an increasing demand for advanced image sensor technologies capable of real-time processing.