ATProto Browser

Record data

{
  "uri": "at://did:plc:x2a3inabvfsn4wntrlbbndrv/app.bsky.feed.post/3llet5st4gf2c",
  "cid": "bafyreibm4rlzmhgrergzikhjsl5zfizhvedrdplxeyxaxyzsh4zh6yjpya",
  "value": {
    "text": "We focus on the *linear softmax* policy parameterization, which is perhaps the simplest concrete policy class. We show that efficient language model exploration has inherent computational-statistical tradeoffs, and clarify the role of the base model in the process.\n\n11/",
    "$type": "app.bsky.feed.post",
    "langs": [
      "en"
    ],
    "reply": {
      "root": {
        "cid": "bafyreidzv7o2gllewwfqqi4ixremjopxmuyntc5gstydqektrxkcjni4pi",
        "uri": "at://did:plc:x2a3inabvfsn4wntrlbbndrv/app.bsky.feed.post/3llet5p66ac2c"
      },
      "parent": {
        "cid": "bafyreia3ionxopl5d4j3pg7wn2hqsd6bysbcf2zr2w5muaqgj3qh464xci",
        "uri": "at://did:plc:x2a3inabvfsn4wntrlbbndrv/app.bsky.feed.post/3llet5st4ge2c"
      }
    },
    "createdAt": "2025-03-27T17:28:13.780Z"
  }
}