arXiv subjects

Dazhi Peng

Publications and source records attributed to Dazhi Peng.

3 recordsLinked to original sources

arXiv2026

ForgeKernel: Self-Evolving Kernels Catching and Surpassing Experts and Agents

Purpose-built kernel agents can write high-performance GPU kernels, but existing systems generally lack cross-run knowledge accumulation and architectural evolution. Each run starts from the same initial knowledge, and later iterations largely optimize within an architecture fixed early in the search. We present ForgeKernel, a kernel-generation system that accumulates and reuses knowledge across runs while enabling continued evolution of kernel architectures. It analyzes the hardware behavior of a high-performance reference and distills validated architectural decisions into reusable knowledge of hardware mechanisms rather than imitating source code. It then seals a kernel that passes correctness and performance validation and uses it as the reference for the next optimization run, enabling architectural decisions to be revisited. Across 27 operators and 29 input configurations on NVIDIA Hopper and Blackwell GPUs, ForgeKernel matches or beats its expert reference on 25 configurations, with a geometric-mean performance gain of 10%. On Ascend 910C, it outperforms the per-operator public best on the public case set of all 8 level-4 operators in the CANN-Bench Season 1 final leaderboard, with a geometric-mean gain of 25.1%, while its full-suite submission ranks first on the A3 leaderboard. As of 2026-10-30, public submissions generated by ForgeKernel rank first on all three CANN-Bench 2.0 solution leaderboards (Ascend 910C, 950DT, and 950PR) and on the SOL-ExecBench FlashInfer-Bench track.

cs.DC↗

arXiv2026

JustGoal: With Only a Goal, the AI Baseline Is Above the Expert Kernels

A small number of GPU kernels often determines the efficiency of large language model training and inference. High-performance kernel development typically relies on expert knowledge through manual implementation, collaboration with general-purpose models, or specialized agents equipped with domain knowledge, evaluation methods, and search workflows. We investigate how much value this knowledge provides as foundation models become stronger. We introduce JustGoal, a goal-only baseline in which a general-purpose agent receives only the optimization objective and constraints, without kernel-specific infrastructure, knowledge bases, or predefined workflows, and builds its own evaluation loop. On NVIDIA B200 GPUs, JustGoal outperforms open-source libraries on 8 of 10 operators, achieving per-operator geometric mean speedups of 1.14x to 2.34x. It also surpasses the fastest publicly released specialized-agent kernel on 5 of 9 operators with available comparisons. In a controlled same-model comparison using Fable 5.1 on single shapes, the KDA-0.5 harness reaches 0.95x of JustGoal's performance, while single-mechanism ablations show no consistent benefit from domain knowledge or predefined workflows. These results indicate that specialized machinery should be justified by clear gains over a same-model, same-scope goal-only baseline and reevaluated whenever the foundation model changes.

cs.DC↗

arXiv2025

X-SQL: Expert Schema Linking and Understanding of Text-to-SQL with Multi-LLMs

With Large Language Models' (LLMs) emergent abilities on code generation tasks, Text-to-SQL has become one of the most popular downstream applications. Despite the strong results of multiple recent LLM-based Text-to-SQL frameworks, the research community often overlooks the importance of database schema information for generating high-quality SQL queries. We find that such schema information plays a significant or even dominant role in the Text-to-SQL task. To tackle this challenge, we propose a novel database schema expert with two components. We first introduce X-Linking, an LLM Supervised Finetuning (SFT)-based method that achieves superior Schema Linking results compared to existing open-source Text-to-SQL methods. In addition, we innovatively propose an X-Admin component that focuses on Schema Understanding by bridging the gap between abstract schema information and the user's natural language question. Aside from better learning with schema information, we experiment with Multi-LLMs for different components within the system to further boost its performance. By incorporating these techniques into our end-to-end framework, X-SQL, we have achieved Execution Accuracies of 84.9% on the Spider-Dev dataset and 82.5% on the Spider-Test dataset. This outstanding performance establishes X-SQL as the leading Text-to-SQL framework based on open-source models.

cs.LG↗