{"id":"5fc975a592c3","page":"/c/5fc975a592c3/","topic":"Internal instruction following eval","publisher":"OpenAI","official":true,"platform":"X","publishedAt":"2025-12-15T21:26:14.000Z","model":"GPT-4o","form":"bar","metrics":{"views":226982,"likes":1036,"reposts":98,"quotes":45,"replies":47,"saves":279},"readAt":"2026-10-13T19:00:27.280Z","chartsInPost":3,"score":71,"scoreWeights":{"views":0.4,"likes":0.3,"eng":0.15,"save":0.15},"quadrant":"wide","ranks":{"score":{"industry":{"n":200,"of":1162},"company":{"n":13,"of":74}},"views":{"industry":{"n":241,"of":1162},"company":{"n":29,"of":74}},"likes":{"industry":{"n":264,"of":1162},"company":{"n":30,"of":74}},"eng":{"industry":{"n":789,"of":1162},"company":{"n":37,"of":74}},"save":{"industry":{"n":322,"of":1162},"company":{"n":6,"of":74}}},"percentiles":{"industry":{"views":0.7930292598967298,"likes":0.7732358003442341,"eng":0.32142857142857145,"save":0.7233218588640276},"company":{"views":0.6148648648648649,"likes":0.6013513513513513,"eng":0.5067567567567568,"save":0.9256756756756757}},"peers":{"industry":1162,"company":74,"companyName":"OpenAI"},"medians":{"industry":{"views":39945,"likes":289,"eng":0.007331788144862199,"save":0.0007642791742968587},"company":{"views":161517.5,"likes":714.5,"eng":0.005359315036129835,"save":0.00048697197603225794},"form":{"n":241,"views":25600,"eng":0.007940930621342993}},"ratios":{"vsCompanyViews":1.4053090222421718,"vsCompanyLikes":1.449965010496851,"viewsPerFollower":0.4964523805357737,"viewsPerDay":751.8480578720855},"followers":457208,"reads":[{"at":"2026-10-12T15:45:44.121Z","views":226981,"likes":1036,"reposts":98,"quotes":45,"replies":47,"saves":279,"source":"x-search-cache"},{"at":"2026-10-12T20:02:43.552Z","views":226981,"likes":1036,"reposts":98,"quotes":45,"replies":47,"saves":279,"source":"x-search-cache"},{"at":"2026-10-12T21:11:11.067Z","views":226981,"likes":1036,"reposts":98,"quotes":45,"replies":47,"saves":279,"source":"x-search-cache"},{"at":"2026-10-13T19:00:27.280Z","views":226982,"likes":1036,"reposts":98,"quotes":45,"replies":47,"saves":279,"source":"bird-read"}],"chartData":"/api/charts/5fc975a592c3.json","copy":{"en":{"why":[{"factor":"rank","text":"Its score of 71 ranks #200 of 1,162 AI chart posts and #13 of 74 OpenAI posts."},{"factor":"multi-chart-post","text":"The post carries 3 charts, and every chart in the post shares the post's numbers."},{"factor":"saves","text":"Its save rate of 0.12% sits at the 72nd percentile among AI chart posts, against a median of 0.076%."}],"verdict":"It sits in the Wide but thin quadrant, at #200 of 1,162 AI chart posts.","news":{"headline":"OpenAI posts internal instruction following eval bar chart","dek":"The December 15, 2025 post drew 227K views and a 0.54% engagement rate, with 3 charts sharing its numbers.","what":"OpenAI posted a bar chart of an internal instruction following eval on December 15, 2025, with 2 models shown. It came with a Common Voice WER comparison and an internal hallucination with noise eval, and it was the account's only post that day.","numbers":"It reached 227K views, ranking #241 of 1,162 AI chart posts against a median of 39.9K. Its score of 71 ranks #200 of 1,162 AI chart posts and #13 of 74 OpenAI posts.","lesson":"A save rate of 0.12% at the 72nd percentile came with an engagement rate of 0.54%: the chart was kept more than it was answered. The 3 charts in the post all share those same numbers.","implication":"At #13 of 74 OpenAI posts, this chart sits well inside the account's own range, while its 0.54% engagement rate falls below the AI chart median. The 227K views against a 39.9K median show the gap between reach and response."}},"zh":{"why":[{"factor":"rank","text":"综合得分71，在1162条AI图表帖中排第200，高于中位水平。"},{"factor":"multi-chart-post","text":"帖子含3张图，每张图共用帖子数据，多图结构未拉低表现。"},{"factor":"saves","text":"收藏率0.12%，高于AI图表帖中位数0.076%，排第72百分位。"}],"verdict":"传得远，互动偏少，综合得分在1162条AI图表帖中排第200。","news":{"headline":"OpenAI内部指令遵循评估图浏览量22.7万","dek":"浏览量高于中位数，互动率低于中位数。","what":"OpenAI于2025年12月15日发布内部指令遵循评估准确率对比图，共3张图，另含Common Voice WER和内部幻觉噪声评估。","numbers":"浏览量22.7万，在1162条AI图表帖中排第241，高于中位数4万。点赞1036，互动率0.54%。","lesson":"收藏率0.12%，在1162条AI图表帖中排第322，高于中位数0.076%。","implication":"在OpenAI的74条帖子中排第13，综合得分71，高于多数自家帖子。"}},"lane":"en:amd/DeepSeek-V4.1-Flash zh:amd/DeepSeek-V4-Flash","checkedAt":"2026-10-13T20:36:43.661Z"}}