<p>
View a PDF of the paper titled <strong>NES: An Instruction-Free, Low-Latency Next Edit Suggestion Framework Powered by Learned Historical Editing Trajectories</strong>, by Xinfang Chen, Siyang Xiao, Xianying Zhu, Junhong Xie, Ming Liang, Dajun Chen, Wei Jiang, Yong Li, and Peng Di.
</p>
<a href="#" title="View PDF">View PDF</a>
<p>
<em>HTML (experimental)</em>
</p>
<blockquote class="abstract mathjax">
<span class="descriptor">Abstract:</span> Code editing is a frequent yet cognitively demanding task in software development. Existing AI-powered tools often disrupt developer flow by requiring explicit natural language instructions and suffer from high latency, limiting real-world usability. We present NES (Next Edit Suggestion), an instruction-free, low-latency code editing framework that leverages learned historical editing trajectories to implicitly capture developers' goals and coding habits. NES features a dual-model architecture: one model predicts the next edit location and the other generates the precise code change, both without any user instruction. Trained on our open-sourced SFT and DAPO datasets, NES achieves state-of-the-art performance (75.6% location accuracy, 27.7% exact match rate) while delivering suggestions in under 250ms. Deployed at Ant Group, NES serves over 20,000 developers through a seamless Tab-key interaction, achieving effective acceptance rates of 51.55% for location predictions and 43.44% for edits, demonstrating its practical impact in real-world development workflows.
</blockquote>
<div>
<h2>Submission History</h2>
From: Peng Di [view email]
<br/>
<strong>[v1]</strong> Mon, 4 Aug 2025 14:37:32 UTC (1,849 KB)<br/>
<strong>[v2]</strong> Tue, 31 Mar 2026 15:41:02 UTC (1,560 KB)<br/>
<strong>[v3]</strong> Wed, 1 Apr 2026 13:01:41 UTC (1,561 KB)<br/>
</div>Understanding the NES Framework: A Game Changer in Code Editing
In the fast-paced world of software development, code editing stands as a critical yet often daunting task. Developers frequently grapple with complex codebases, and the mental strain it imposes can disrupt their workflow. The introduction of AI-powered tools has promised to enhance coding efficiency, but often, these tools come with their own set of challenges. Many require explicit natural language instructions and introduce high latency, which can be frustrating and counterproductive.
Enter NES (Next Edit Suggestion), a revolutionary framework that aims to redefine how developers interact with code editors. By employing learned historical editing trajectories, NES bypasses the need for explicit instructions, allowing it to provide seamless and contextually appropriate suggestions.
How NES Works: A Dual-Model Architecture
At the heart of NES lies an innovative dual-model architecture. The first model is designed to predict the next edit location—where a developer is likely to make changes in their code. The second model focuses on generating the precise changes needed, all without requiring any input or guidance from the user. This unique approach not only decreases the cognitive load on developers but also allows for a more intuitive and fluid coding experience.
Impressive Performance Metrics
One of the standout features of NES is its remarkable performance metrics. Trained on comprehensive datasets like SFT and DAPO, NES boasts an accuracy rate of 75.6% for location predictions and a 27.7% exact match rate for code changes. Moreover, it operates with an impressive response time of under 250 milliseconds, which significantly enhances the tool’s usability in real-world development scenarios.
Real-World Impact: Deployment at Ant Group
NES isn’t merely a theoretical concept; it has been deployed at Ant Group, where it serves over 20,000 developers. This widespread implementation highlights its practical utility and the trust placed in its capabilities. By leveraging a seamless Tab-key interaction, NES aligns with the natural workflow of developers, allowing them to accept suggestions effortlessly. According to current acceptance rates, 51.55% of location predictions and 43.44% of edit suggestions are accepted, illustrating NES’s effectiveness in enhancing productivity.
More Read
![Efficient Reasoning Through Discounted Reinforcement Learning: Insights from Paper [2510.23486]](https://aimodelkit.com/wp-content/uploads/2026/07/1783458025_arxiv-logo-fb-150x150.png "Efficient Reasoning Through Discounted Reinforcement Learning: Insights From Paper [2510.23486]"){kind=logo}
The Benefits of an Instruction-Free Framework
The instruction-free nature of NES is one of its most compelling features. Traditional AI tools often disrupt a developer’s flow by requiring explicit commands, which can be cumbersome and time-consuming. NES eliminates this barrier, allowing developers to focus on writing code rather than on articulating their intentions. By understanding and predicting developers’ goals based on their past editing behavior, NES facilitates a more organic coding experience.
Future Implications for Software Development
As more organizations begin to adopt AI-driven solutions like NES, the implications for software development could be profound. Enhanced productivity, reduced cognitive load, and improved developer satisfaction are just a few potential outcomes. The ability of such frameworks to learn and adapt over time means that they can continue to evolve alongside the developers using them, creating a more tailored and effective coding environment.
In summary, NES represents a significant advancement in code editing technology, positioned to transform how developers interact with their code. By providing low-latency, instruction-free suggestions, it not only increases efficiency but also enriches the overall experience of software development. As this technology continues to mature, it may well set a new standard for the tools of the trade, guiding a new era of seamless coding.
Inspired by: Source
