ATProto Browser

Record data

{
  "uri": "at://did:plc:x2a3inabvfsn4wntrlbbndrv/app.bsky.feed.post/3lobv4ecsl22d",
  "cid": "bafyreidtqsd3leeprtf22li2xtwmtenzimovupduyixcicvsq3ksofhs44",
  "value": {
    "text": "Inference-time computation can boost language model performance, but scaling naively—like with Best-of-N sampling (selecting the top-scoring response from multiple candidates with a reward model)—can actually degrade results through reward hacking (eg, this fig from arxiv.org/abs/2210.10760).\n\n2/11",
    "$type": "app.bsky.feed.post",
    "embed": {
      "$type": "app.bsky.embed.images",
      "images": [
        {
          "alt": "",
          "image": {
            "$type": "blob",
            "ref": {
              "$link": "bafkreifvajumenlmrurnytp6br4lqbthlnbm43udhg3wawkuvohgflq4dm"
            },
            "mimeType": "image/jpeg",
            "size": 251422
          },
          "aspectRatio": {
            "width": 984,
            "height": 710
          }
        }
      ]
    },
    "langs": [
      "en"
    ],
    "reply": {
      "root": {
        "cid": "bafyreih6mvnxmoz4bgad7vcbv2fl63qhwwggtht5lpckbzj7yexbnz2qie",
        "uri": "at://did:plc:x2a3inabvfsn4wntrlbbndrv/app.bsky.feed.post/3lobv4byuec2d"
      },
      "parent": {
        "cid": "bafyreih6mvnxmoz4bgad7vcbv2fl63qhwwggtht5lpckbzj7yexbnz2qie",
        "uri": "at://did:plc:x2a3inabvfsn4wntrlbbndrv/app.bsky.feed.post/3lobv4byuec2d"
      }
    },
    "facets": [
      {
        "index": {
          "byteEnd": 295,
          "byteStart": 271
        },
        "features": [
          {
            "uri": "https://arxiv.org/abs/2210.10760",
            "$type": "app.bsky.richtext.facet#link"
          }
        ]
      }
    ],
    "createdAt": "2025-05-03T17:40:49.560Z"
  }
}