# To Add Is Machine, To Delete Is Human: Measuring and Mitigating Deletion Avoidance in LLM Code Editing

arXiv: https://arxiv.org/abs/2607.28887 · HTML: https://mlsanity.com/paper/2607.28887/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 14 of its 28 references that are in the map.
- First posted on arXiv 30 Jul 2026.

## Builds on (most cited first, 14 of 14)

- [SWE-bench: Can Language Models Resolve Real-World GitHub Issues?](https://mlsanity.com/paper/2310.06770/) (2023)
- [Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces](https://mlsanity.com/paper/2601.11868/) (2026)
- [SWE-Bench+: Enhanced Coding Benchmark for LLMs](https://mlsanity.com/paper/2410.06992/) (2024)
- [LatestEval: Addressing Data Contamination in Language Model Evaluation through Dynamic and Time-Sensitive Test Construction](https://mlsanity.com/paper/2312.12343/) (2023)
- [Can It Edit? Evaluating the Ability of Large Language Models to Follow Code Editing Instructions](https://mlsanity.com/paper/2312.12450/) (2023)
- [GitTaskBench: A Benchmark for Code Agents Solving Real-World Tasks Through Code Repository Leveraging](https://mlsanity.com/paper/2508.18993/) (2025)
- [DOMAINEVAL: An Auto-Constructed Benchmark for Multi-Domain Code Generation](https://mlsanity.com/paper/2408.13204/) (2024)
- [DSCodeBench: A Realistic Benchmark for Data Science Code Generation](https://mlsanity.com/paper/2505.15621/) (2025)
- [EDIT-Bench: Evaluating LLM Abilities to Perform Real-World Instructed Code Edits](https://mlsanity.com/paper/2511.04486/) (2025)
- [Coding Agents Don't Know When to Act](https://mlsanity.com/paper/2605.07769/) (2026)
- [What to Cut? Predicting Unnecessary Methods in Agentic Code Generation](https://mlsanity.com/paper/2602.17091/) (2026)
- [Building to the Test: Coding Agents Deliver What You Check, Not What You Requested](https://mlsanity.com/paper/2606.28430/) (2026)
- [Why Are AI Agent Involved Pull Requests (Fix-Related) Remain Unmerged? An Empirical Study](https://mlsanity.com/paper/2602.00164/) (2026)
- [Loc2Repair: A Framework for Evaluating the Impact of File-Level Issue Localization in Repo-Level LLM Repair](https://mlsanity.com/paper/2606.30963/) (2026)

## Built on by (most cited first, 1 of 1)

- ["Nothing to See Here'': Unintended Disclosure through Revision Traces of LLM Deliverables](https://mlsanity.com/paper/2609.35408/) (2026)

## Abstract (arXiv)

Large language models increasingly write and repair production code, yet evidence is mounting that their test-passing patches leave codebases harder to maintain. We identify one concrete source: deletion avoidance, the systematic tendency to retain code that an intended edit requires removing. Across the five leading models on the official SWE-bench Verified leaderboard, deletion recall against the developer patch reaches at most 71.7% even on tasks all five solve, and models reach the right file for over 92% of required deletions but cut the exact line in under 52% of cases. Instead, 29.0% of passing patches wrap the targeted code in a guard or fallback, a pattern we call Guard-and-Go. Such patches pass because the original tests rarely check removal: when we retrofit 34 Verified tasks with tests that fail if the targeted code remains, four frontier models spanning closed and open weights fall from 63.2% to 41.9%. Because real repairs mix removal with addition, we curate CanItDelete, a benchmark of 200 tasks mined from real commits whose entire required edit is deletion. Even with the addition work gone, the best model still fails one task in five, and smaller open models fall to 18.0%. We then ablate GPT-5.6 Sol under four cumulative prompts; success moves little until we supply the exact lines, which nearly eliminate incomplete deletion yet raise success only to 80.5% because the model then deletes beyond the spans or adds code instead. Finally, through a pilot study we show one potential fix: teaching deletion during post-training reduces deletion avoidance and improves broader code-editing performance, suggesting the behavior is undertrained rather than beyond reach.