AI Paper Atlas

Research map / Large language models / Agents, code and math

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

Wayne Chi, Valerie Chen, Ryan Shar, Aditya Mittal, Jenny Liang, Wei-Lin Chiang, Anastasios Nikolas Angelopoulos, Ion Stoica, Graham Neubig, Ameet Talwalkar, Chris Donahue

ICLR 2026 · Large language models › Agents, code and math · approach: Software engineering agents

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add highlighted code to the prompt (ablation)top model of each of 7 families · EDIT-Bench corepass@1 (%): claude-sonnet-4 62.41 -> 64.81; o3-mini 60.00 -> 56.30Mixed
add cursor position on top of highlighted code (ablation)top model of each of 7 families · EDIT-Bench corepass@1 (%): glm-4.6 56.48 -> 44.81; kimi-k2-0905 56.48 -> 58.15Mixed
add highlighted code for o3-mini (ablation)o3-mini · EDIT-Bench corepass@1 (%): 60.00 -> 56.30Hurt
add highlighted code for glm-4.6 (ablation)glm-4.6 · EDIT-Bench corepass@1 (%): 52.96 -> 56.48Improved
add cursor position for o3-mini and qwen3-coder (ablation)o3-mini, qwen3-coder · EDIT-Bench corepass@1 (%): o3-mini 56.30 -> 55.19; qwen3-coder 53.89 -> 53.89No clear effect
higher reasoning effort or default reasoning in gpt-5 vs gpt-5-minigpt-5 · EDIT-Bench corepass@1Hurt
use a fully agentic pipeline to generate test harnesses instead of human annotatorstest harness creation · EDIT-Benchpass@1 (%): 0%Hurt
compare easy vs hard problems across models40 evaluated LLMs · EDIT-Benchpass@1 gap: average gap 58.8%Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy