EDIT-Bench: Evaluating LLM Abilities to Perform Real-World Instructed Code Edits

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chi, Wayne, Chen, Valerie, Shar, Ryan, Mittal, Aditya, Liang, Jenny, Chiang, Wei-Lin, Angelopoulos, Anastasios Nikolas, Stoica, Ion, Neubig, Graham, Talwalkar, Ameet, Donahue, Chris
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912714801020928
author Chi, Wayne
Chen, Valerie
Shar, Ryan
Mittal, Aditya
Liang, Jenny
Chiang, Wei-Lin
Angelopoulos, Anastasios Nikolas
Stoica, Ion
Neubig, Graham
Talwalkar, Ameet
Donahue, Chris
author_facet Chi, Wayne
Chen, Valerie
Shar, Ryan
Mittal, Aditya
Liang, Jenny
Chiang, Wei-Lin
Angelopoulos, Anastasios Nikolas
Stoica, Ion
Neubig, Graham
Talwalkar, Ameet
Donahue, Chris
contents Instructed code editing, where LLMs directly modify a developer's existing code based on a user instruction, is becoming a widely used interaction mode in AI coding assistants. However, few benchmarks directly evaluate this capability and current datasets often rely on artificial sources. We introduce EDIT-Bench, a benchmark for evaluating LLM code editing capabilities grounded in real-world usage, i.e., user instructions and code contexts collected in the wild. EDIT-Bench comprises of 540 problems, multiple natural and programming languages, and a diverse set of real-world use cases, ranging from resolving errors to adding features. EDIT-Bench introduces context-dependent problems that require the model to understand code context, highlighted code, and cursor position in addition to the user instruction. We evaluate 40 diverse LLMs and observe that EDIT-Bench is a challenging set of problems where only 1 model scores over 60%. We find that model performance varies across different categories of user instructions. Further, we find that varying levels of contextual information greatly affect task success rate, with performance varying up to 11%, indicating the importance of evaluating with realistic context.
format Preprint
id arxiv_https___arxiv_org_abs_2511_04486
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle EDIT-Bench: Evaluating LLM Abilities to Perform Real-World Instructed Code Edits
Chi, Wayne
Chen, Valerie
Shar, Ryan
Mittal, Aditya
Liang, Jenny
Chiang, Wei-Lin
Angelopoulos, Anastasios Nikolas
Stoica, Ion
Neubig, Graham
Talwalkar, Ameet
Donahue, Chris
Software Engineering
Instructed code editing, where LLMs directly modify a developer's existing code based on a user instruction, is becoming a widely used interaction mode in AI coding assistants. However, few benchmarks directly evaluate this capability and current datasets often rely on artificial sources. We introduce EDIT-Bench, a benchmark for evaluating LLM code editing capabilities grounded in real-world usage, i.e., user instructions and code contexts collected in the wild. EDIT-Bench comprises of 540 problems, multiple natural and programming languages, and a diverse set of real-world use cases, ranging from resolving errors to adding features. EDIT-Bench introduces context-dependent problems that require the model to understand code context, highlighted code, and cursor position in addition to the user instruction. We evaluate 40 diverse LLMs and observe that EDIT-Bench is a challenging set of problems where only 1 model scores over 60%. We find that model performance varies across different categories of user instructions. Further, we find that varying levels of contextual information greatly affect task success rate, with performance varying up to 11%, indicating the importance of evaluating with realistic context.
title EDIT-Bench: Evaluating LLM Abilities to Perform Real-World Instructed Code Edits
topic Software Engineering
url https://arxiv.org/abs/2511.04486