[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-a-new-ai-model-learns-to-read-the-night-sky":10,"sections":34},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":24,"tags":25,"sources":29,"feedback":33,"feedback_at":22,"cost_usd":33,"total_tokens":33},2583,"a-new-ai-model-learns-to-read-the-night-sky","A New AI Model Learns to Read the Night Sky","Researchers built a light-curve model that handles astronomy's messy, irregular data better than hand-crafted features on 15 of 16 benchmarks.","A self-supervised AI framework trained on astronomical brightness data now outperforms traditional feature engineering across nearly every tested metric.\n\nAstronomers track how stars and other objects brighten or dim over time using what are called light curves. The problem: telescopes don't observe on a neat schedule, and the data arrives unevenly sampled, noisy, and spanning wildly different physical timescales — exactly the conditions that trip up most off-the-shelf time-series models. Researchers addressed this with a framework built on a Joint-Embedding Predictive Architecture, or JEPA, layering in uncertainty-aware tokenization and a training method called multi-view self-distillation. Tested on the StarEmbed classification benchmark, the model beat hand-crafted features on 15 of 16 metrics. In few-shot settings — the hardest test, where the model sees very little labeled data — it reached a macro-F1 score of 42.56 with just one labeled example per class, climbing to 63.58 with 100.\n\nThe broader claim here is that domain knowledge baked into the architecture matters more than scale alone. That's a pointed rebuke of the \"one model to rule them all\" instinct driving much of current AI research. The framework also transferred reasonably well to 12 unrelated irregular time-series datasets, matching or beating prior state-of-the-art on five of them — more wins than any single previous baseline managed.\n\nThe results are a reminder that foundation models built for general text or even general time-series data still stumble when physics gets in the way — and that a carefully constrained model often beats a larger, blunter one.","[\"ai\",\"science\",\"machine-learning\",\"astronomy\"]","2026-06-30T04:00:00.000Z","2026-06-30T08:37:26.457Z","2026-06-30T08:37:29.361Z","published",null,[],"science",[26,24,27,28],"ai","machine-learning","astronomy",[30],{"name":31,"url":32},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2606.28446",0,{"sections":35},[36,40,45,50,55,60,65,70,74,79,84,88,93,98],{"name":37,"slug":26,"count":38,"latest_published_at":39},"AI",2590,"2026-07-16T04:00:00.000Z",{"name":41,"slug":42,"count":43,"latest_published_at":44},"Security","security",294,"2026-07-15T19:59:48.000Z",{"name":46,"slug":47,"count":48,"latest_published_at":49},"Deals","deals",179,"2026-06-29T20:02:07.000Z",{"name":51,"slug":52,"count":53,"latest_published_at":54},"Policy","policy",158,"2026-07-16T00:02:48.000Z",{"name":56,"slug":57,"count":58,"latest_published_at":59},"Hardware","hardware",122,"2026-07-14T19:46:26.000Z",{"name":61,"slug":62,"count":63,"latest_published_at":64},"Consumer Tech","consumer-tech",93,"2026-07-13T13:20:48.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":69},"Software","software",70,"2026-07-13T19:52:25.000Z",{"name":71,"slug":24,"count":72,"latest_published_at":73},"Science",66,"2026-07-10T10:29:37.000Z",{"name":75,"slug":76,"count":77,"latest_published_at":78},"Dev Tools","dev-tools",59,"2026-07-07T04:00:00.000Z",{"name":80,"slug":81,"count":82,"latest_published_at":83},"Gaming","gaming",41,"2026-07-09T04:00:00.000Z",{"name":85,"slug":86,"count":82,"latest_published_at":87},"Startups","startups","2026-06-29T20:55:50.000Z",{"name":89,"slug":90,"count":91,"latest_published_at":92},"General","general",29,"2026-07-10T22:28:58.000Z",{"name":94,"slug":95,"count":96,"latest_published_at":97},"Reviews","reviews",20,"2026-06-24T12:00:01.000Z",{"name":99,"slug":100,"count":101,"latest_published_at":102},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]