Researchllmvulnerability patchingjavasecurity
LLMs Show Limited Vulnerability Patching Ability
6.0
On Dec. 11, 2025, researchers tested LLMs from OpenAI, Meta, DeepSeek, and Mistral to see if they could automatically fix vulnerable Java functions in a single attempt. The experiments evaluated two vulnerability groups and found inconsistent success: models repaired some bugs but often produced incorrect or incomplete patches. The results suggest LLMs can assist but require human review and specialized tooling for reliable patching.
Key Points
- 1Tested LLMs from OpenAI, Meta, DeepSeek, Mistral on fixing vulnerable Java functions in one attempt
- 2Found models succeeded inconsistently, showing strengths on some vulnerability types but failing on others
- 3Indicates practitioners need human review and targeted tooling; LLM outputs are not yet reliable patches
Scoring Rationale
Notable study across multiple LLMs gives practical insight, but limited novelty and single-study scope reduce impact.
Sources
Public references used for this report.
Practice with real Logistics & Shipping data
90 SQL & Python problems · 15 industry datasets
Used by DS/ML engineers at top companies
High-Value Overnight OrdersEasyDelivered International ShipmentsMediumOn-Time Delivery Rate by CarrierHard
250 free problems · No credit card
See all Logistics & Shipping problems
