---
format: "aidr-story-markdown/v1"
id: "7e4662fd94a38d37106a915c19ec39f01efe3688bc5937e9c1ba39ca42fadead"
canonical_url: "https://aidr.today/7e4662fd?lang=en"
title: "AscendGYM: A Cost-Aware Kernel Optimization Benchmark for Real-World Workloads on Ascend NPUs"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-11-18T00:00:00.000Z"
category: "Open Source"
topics: ["llm","open-source"]
source_urls: ["https://huggingface.co/papers/2611.17621"]
summary: "Kernel optimization is essential for efficient large language model (LLM) training and inference on Ascend NPUs. Existing Ascend kernel benchmarks primarily focus on kernel generation, with less emphasis on the ability of agents to optimize existing implementations under input configurations drawn from real workloads. To address this requirement, we introduce AscendGYM, a benchmark whose initial suite comprises 48 kernels across three difficulty levels. Each kernel is paired with test cases whose data types and tensor shapes are collected from real-world training and inference tasks. Consequently, performance improvements on AscendGYM map directly to the corresponding workloads. In terms of usability, two evaluation modes are supported in AscendGYM: model evaluation and model-harness evaluation. The kernel optimization capabilities of both trained models and custom agents are allowed to be measured. It provides a complete execution environment that closely reproduces how Ascend kernels are invoked in practical pipelines. For model evaluation, it additionally includes a built-in harness equipped with search orchestration and experience accumulation."
---

# AscendGYM: A Cost\-Aware Kernel Optimization Benchmark for Real\-World Workloads on Ascend NPUs

> [Open the canonical story](<https://aidr.today/7e4662fd?lang=en>)

**Published:** 2026-11-18T00:00:00.000Z
**Category:** Open Source
**Topics:** llm, open\-source

## Summary

Kernel optimization is essential for efficient large language model \(LLM\) training and inference on Ascend NPUs\. Existing Ascend kernel benchmarks primarily focus on kernel generation, with less emphasis on the ability of agents to optimize existing implementations under input configurations drawn from real workloads\. To address this requirement, we introduce AscendGYM, a benchmark whose initial suite comprises 48 kernels across three difficulty levels\. Each kernel is paired with test cases whose data types and tensor shapes are collected from real\-world training and inference tasks\. Consequently, performance improvements on AscendGYM map directly to the corresponding workloads\. In terms of usability, two evaluation modes are supported in AscendGYM: model evaluation and model\-harness evaluation\. The kernel optimization capabilities of both trained models and custom agents are allowed to be measured\. It provides a complete execution environment that closely reproduces how Ascend kernels are invoked in practical pipelines\. For model evaluation, it additionally includes a built\-in harness equipped with search orchestration and experience accumulation\.

## Sources

- [Story source](<https://huggingface.co/papers/2611.17621>)

