Experimental browser for the Atmosphere
To better understand this, we study next-token prediction in misspecified setting, where the goal is to learn as well as the best-in-class model up to an approximation factor C≥1. 4/11
Feb 23, 2025, 9:00 PM
{ "uri": "at://did:plc:x2a3inabvfsn4wntrlbbndrv/app.bsky.feed.post/3liuq3mrwks2e", "cid": "bafyreigmk2kaj56ubcjeamp4uvbysvdq5uac63gf3i67h7i6sbkaskyjpe", "value": { "text": "To better understand this, we study next-token prediction in misspecified setting, where the goal is to learn as well as the best-in-class model up to an approximation factor C≥1.\n\n4/11", "$type": "app.bsky.feed.post", "langs": [ "en" ], "reply": { "root": { "cid": "bafyreids3t5hbin5gzohv5ftcospo6v4izvleqfb2uzygkbc7dok3yncuq", "uri": "at://did:plc:x2a3inabvfsn4wntrlbbndrv/app.bsky.feed.post/3liuq3jtnxk2e" }, "parent": { "cid": "bafyreifs7o7b6h2bu23ot5mjccqitsdor572fkvj6kjto4thv3pofr6u24", "uri": "at://did:plc:x2a3inabvfsn4wntrlbbndrv/app.bsky.feed.post/3liuq3mchh22e" } }, "createdAt": "2025-02-23T21:00:20.783Z" } }