Memdoor's multi-agent coding reproduction beats Fable 5 in a ten-problem test
What happened
On November 1, Hacker News' front page covered Memdoor's reproduction of the GVS5H multi-agent pipeline. The test ran five Qwen3.8-27B models over five rounds on the first ten LiveCodeBench Hard problems, solving 9.8 on average. The reported comparisons: the original author's pipeline averaged 9.6, and the single-model Fable 5 averaged 8.0. On this ten-problem test, Memdoor's reproduction solved more on average than both. The report gives no results on a larger problem set.
Written by AI from the coverage · updated 1 hour ago
Coverage
Follow the reports to see the story from different sides.
- Hacker News front pageMemdoor: A team of 5 Qwen3.8-27B models beats Fable 5 on LiveCodeBench Hard
Memdoor 复现 GVS5H 多智能体流程，五个 Qwen3.8-27B 在前十道 LiveCodeBench Hard 题目上测试五轮，平均解出 9.8 题，原作者流程为 9.6 题、Fable 5 单模型为 8.0 题。
Heat over time
Not enough continuous observations to draw a trend yet.