/

← All benchmarks

RepoBench

coding · repo-completion↗ Quelle

55Evidence

active

What it measures

Codevervollstaendigung mit Kontext aus dem umgebenden Repository.

What it does not measure

Agentisches Vorgehen, Testausfuehrung.

How the score is composed

Currency94

Discrimination0

Headroom100

Contamination safety55

Gaming resistance55

Methodological rigor50

Reproducibility60

Details

Benchmaxxing risk: medium
Difficulty: medium
Public test set: yes
Human-verified: no
LLM-judged: no
Letzter Messwert: 2026-10-21
Skala: 0–100 % exact/edit (höher besser)

Best measured models

1ERNIE-5.5-FlashBaiduvendor number49.1 % exact/edit