~/home / Hacker News
Hacker News · Show HN

I audited my AI leaderboard scale – every score dropped 6-15 points

一位开发者自测AI排行榜评分，发现所有模型分数均下降6-15分，揭示了评估基准可能存在的系统性偏差问题
▲41/天
为什么值得关注

揭示了当前AI评测体系的脆弱性，推动更严谨的基准测试方法论

AI评估开发者工具
访问 Hacker News 页面 →
分享到 X
手机端点「分享」直达微信/朋友圈/小红书;桌面端用「复制文案」后到 App 内粘贴发布

同类产品

Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series MacCheapFoodMap – A map of good meals under $10Qwen Scribe – local transcription and dictation for Apple SiliconKedge – Full-stack cloud with forkable VM snapshots and global SQLiteI was tired of opening 2 tabs for every HN link, so I made a userscriptDreeve, a self-hosted dashboard for your sports and fitness data