---
格式版本: 2
标题: "GRADSOLVE: fast exact gradients for ODE ensembles on GPUs"
原文链接: "https://arxiv.org/abs/2609.02876"
发布日期: "2026-09-02"
发布时间校准状态: "found"
发布时间需复核: "否"
发布时间来源: "rule:local:strict_original_body"
发布时间证据: "**\\[v1\\]** Wed, 2 Sep 2026 17:56:10 UTC (252 KB)"
发布时间校准原因: "规则确认唯一严格发布时间，来源 local:strict_original_body"
发布时间校准置信度: "high"
发布时间候选数量: 18
发布时间严格候选数量: 6
发布时间原页读取状态: "source template page reused from URL open"
发布时间未找到原因: ""
发布时间校准时间: "2026-09-04T01:33:36+08:00"
发布时间仲裁状态: "skipped"
发布时间仲裁尝试次数: 0
发布时间仲裁耗时毫秒: 0
发现时间: "2026-09-04T01:32:14+08:00"
入库时间: "2026-09-03T17:33:36.570Z"
来源平台: "arXiv 学术论文搜索"
搜索渠道: "source_template"
搜索词: "https://arxiv.org/search/?query=GPU&searchtype=all"
匹配关键词:
  - "GPU"
相关厂家:
  - "NVIDIA"
相关专家:
  []
内容类型: "网页"
抓取工具: "Free Fetch + Defuddle"
清洗工具: "Defuddle Markdown + Defuddle/Readability 正文提取"
原始附件:
  []
AI优质: "否"
AI打分: 19
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "论文研究GPU上ODE数值微分，与超节点/AI Rack等主题无关，不涉及硬件架构、互连或部署。"
AI质检模型: "tx-deepseek-v4-flash"
AI质检时间: "2026-09-04T01:34:29+08:00"
AI主题相关性: 1
AI来源权威性: 3
AI新颖性: 5
AI技术细节: 3
AI商业部署信号: 0
AI完整性: 7
AI摘要: "GRADSOLVE 是一个开源的 JAX 库，用于在 NVIDIA GPU 上求解并反向微分低维常微分方程（ODE）集成，提供精确离散伴随梯度。"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-09-04T00:18:15.398Z"
采集批次: "2026年9月3日22点43分34秒"
采集批次ID: "20260903-224334-406"
去重键: "https://arxiv.org/abs/2609.02876"
---

## Computer Science > Mathematical Software

## Title:GRADSOLVE: fast exact gradients for ODE ensembles on GPUs

Authors:[Alessio Spurio Mancini](https://arxiv.org/search/cs?searchtype=author&query=Mancini,+A+S)

[View PDF](https://arxiv.org/pdf/2609.02876) [HTML (experimental)](https://arxiv.org/html/2609.02876v1)

> Abstract:Ordinary differential equations (ODEs) underlie models in science and engineering, and many applications need derivatives of their solutions with respect to parameters. Ensembles of independent trajectories suit graphics processing units (GPUs), but current GPU software forces a trade-off: the fastest ensemble solvers cannot be differentiated in reverse mode at the speed they solve, and the solvers built for differentiation solve more slowly. No single tool has yet offered a reverse-mode gradient at the speed of a fused-kernel solve.  
> We present GRADSOLVE, an open-source JAX library for solving and reverse-mode differentiating low-dimensional ODE ensembles on NVIDIA GPUs. It records the steps an adaptive solver accepts and differentiates a fixed-step replay of them; the returned gradient is the exact discrete adjoint of those steps, the same derivative Diffrax returns by default, obtained more cheaply from a fixed-length chain than from an adaptive loop. It targets ensembles differentiated many times against one recorded mesh, keeps Diffrax as a fallback, and supports explicit and Rosenbrock integrators.  
> Used as a solver, GRADSOLVE's forward-only kernel ran 2.8x faster than [this http URL](http://diffeqgpu.jl/); used for gradients, once a record exists, it computed them 5.6-14.1x faster than Diffrax's checkpointed adjoint at matched forward-state accuracy across three GPU generations, the advantage narrowing on large ensembles and, on stiff systems, down to parity at tight accuracy. GRADSOLVE is released at [this https URL](https://github.com/ECLIPSE-AI4Science/gradsolve).

| Comments: |  |
| --- | --- |
| Subjects: | Mathematical Software (cs.MS); Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG); Numerical Analysis (math.NA) |
| Cite as: | [arXiv:2609.02876](https://arxiv.org/abs/2609.02876) \[cs.MS\] |
|  | (or [arXiv:2609.02876v1](https://arxiv.org/abs/2609.02876v1) \[cs.MS\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2609.02876](https://doi.org/10.48550/arXiv.2609.02876) |

## Submission history

From: Alessio Spurio Mancini Dr. \[[view email](https://arxiv.org/show-email/fbb2fcbb/2609.02876)\]  
**\[v1\]** Wed, 2 Sep 2026 17:56:10 UTC (252 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2609.02876) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))