ATProto Browser

Record data

{
  "uri": "at://did:plc:x2a3inabvfsn4wntrlbbndrv/app.bsky.feed.post/3linhl7u7f227",
  "cid": "bafyreidgrnshhudokp53dd4ulmmdyqqezeuirt6phs3miz2nptv7uk6rwy",
  "value": {
    "text": "Result 2: If resetting to previously observed states is allowed (a-la arxiv.org/abs/2404.15417), one can learn efficiently using only one-state regression instead of two-state, bypassing the hardness of [GMR’24] and giving a computational separation between fully online RL and RL with resets.\n\n8/",
    "$type": "app.bsky.feed.post",
    "langs": [
      "en"
    ],
    "reply": {
      "root": {
        "cid": "bafyreih52si2tydkay3bwofyxoa6oc52b6t2ymopfurvsdn24artsfeim4",
        "uri": "at://did:plc:x2a3inabvfsn4wntrlbbndrv/app.bsky.feed.post/3linhl4yguc27"
      },
      "parent": {
        "cid": "bafyreifudw24bjo5pkgiougvkfdrldwmeullqoexfvyitb3dqkmrhfhebi",
        "uri": "at://did:plc:x2a3inabvfsn4wntrlbbndrv/app.bsky.feed.post/3linhl7u6fs27"
      }
    },
    "facets": [
      {
        "index": {
          "byteEnd": 94,
          "byteStart": 70
        },
        "features": [
          {
            "uri": "https://arxiv.org/abs/2404.15417",
            "$type": "app.bsky.richtext.facet#link"
          }
        ]
      }
    ],
    "createdAt": "2025-02-20T23:39:22.337Z"
  }
}