🐦 Twitter Post Details

Viewing enriched Twitter post

@iScienceLuvr

Demystifying Reinforcement Learning Post-Training of Language Models A very useful exploration of RLVR in a simplified setup, lots of good insights... "Result 1: sparse RL cannot find what the model never samples" "Result 2: dense rewards break that ceiling" "Result 3: "spurious rewards" are a story about the prompt set" "Post-training is best understood as redistributing probability mass inside the pretrained distribution. That framing suggests the measurement: track the probability the model assigns to the behavior you want, and track the entropy of its output distribution, throughout training." project page: https://t.co/iPL9jqlxEL code: https://t.co/K9MtdC28Sl link: https://t.co/RQfH9TqfLy

Media 1
View on Twitter

📊 Media Metadata

{
  "media": [
    {
      "type": "photo",
      "url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2092900877031895511/media_0.jpg",
      "filename": "media_0.jpg"
    }
  ],
  "processed_at": "2026-08-27T10:02:07.156412",
  "pipeline_version": "2.0"
}

🔧 Raw API Response

{
  "type": "tweet",
  "id": "2092900877031895511",
  "url": "https://x.com/iScienceLuvr/status/2092900877031895511",
  "twitterUrl": "https://twitter.com/iScienceLuvr/status/2092900877031895511",
  "text": "Demystifying Reinforcement Learning Post-Training of Language Models\n\nA very useful exploration of RLVR in a simplified setup, lots of good insights...\n\n\"Result 1: sparse RL cannot find what the model never samples\"\n\n\"Result 2: dense rewards break that ceiling\"\n\n\"Result 3: \"spurious rewards\" are a story about the prompt set\"\n\n\"Post-training is best understood as redistributing probability mass inside the pretrained distribution. That framing suggests the measurement: track the probability the model assigns to the behavior you want, and track the entropy of its output distribution, throughout training.\"\n\nproject page: https://t.co/iPL9jqlxEL\ncode: https://t.co/K9MtdC28Sl\nlink: https://t.co/RQfH9TqfLy",
  "source": "Twitter for iPhone",
  "retweetCount": 6,
  "replyCount": 1,
  "likeCount": 24,
  "quoteCount": 2,
  "viewCount": 2301,
  "createdAt": "Thu Aug 27 09:03:47 +0000 2026",
  "lang": "en",
  "bookmarkCount": 26,
  "isReply": false,
  "inReplyToId": null,
  "conversationId": "2092900877031895511",
  "displayTextRange": [
    0,
    273
  ],
  "inReplyToUserId": null,
  "inReplyToUsername": null,
  "author": {
    "type": "user",
    "userName": "iScienceLuvr",
    "url": "https://x.com/iScienceLuvr",
    "twitterUrl": "https://twitter.com/iScienceLuvr",
    "id": "441465751",
    "name": "Tanishq Mathew Abraham, Ph.D.",
    "isVerified": false,
    "isBlueVerified": true,
    "verifiedType": null,
    "profilePicture": "https://pbs.twimg.com/profile_images/2058414948255211520/8jVMerrR_normal.jpg",
    "coverPicture": "https://pbs.twimg.com/profile_banners/441465751/1779599917",
    "description": "",
    "location": "",
    "followers": 90245,
    "following": 1525,
    "status": "",
    "canDm": true,
    "canMediaTag": false,
    "createdAt": "Tue Dec 20 03:45:50 +0000 2011",
    "entities": {
      "description": {
        "urls": []
      },
      "url": {}
    },
    "fastFollowersCount": 0,
    "favouritesCount": 131455,
    "hasCustomTimelines": true,
    "isTranslator": false,
    "mediaCount": 2984,
    "statusesCount": 21368,
    "withheldInCountries": [],
    "affiliatesHighlightedLabel": {},
    "possiblySensitive": false,
    "pinnedTweetIds": [
      "2088229088322560106"
    ],
    "profile_bio": {
      "description": "CEO @SophontAI |\nFounder @MedARC_AI |\nPhD at 19 (2023) |\nex Research Director Stability AI | \nBiomed. engineer @ 14 |\nTEDx talk➡https://t.co/xPxwKTq6Qb",
      "entities": {
        "description": {
          "urls": [
            {
              "display_url": "bit.ly/3tpAuan",
              "expanded_url": "https://bit.ly/3tpAuan",
              "indices": [
                128,
                151
              ],
              "url": "https://t.co/xPxwKTq6Qb"
            }
          ]
        },
        "url": {
          "urls": [
            {
              "display_url": "sophont.med",
              "expanded_url": "https://sophont.med",
              "indices": [
                0,
                23
              ],
              "url": "https://t.co/MvROZZW1Zg"
            }
          ]
        }
      }
    },
    "isAutomated": false,
    "automatedBy": null
  },
  "extendedEntities": {
    "media": [
      {
        "allow_download_status": {
          "allow_download": true
        },
        "display_url": "pic.x.com/vE4jOrvrq1",
        "expanded_url": "https://x.com/iScienceLuvr/status/2092900877031895511/photo/1",
        "ext_media_availability": {
          "status": "Available"
        },
        "features": {
          "large": {
            "faces": []
          },
          "medium": {
            "faces": []
          },
          "orig": {
            "faces": []
          },
          "small": {
            "faces": []
          }
        },
        "id_str": "2092900819368595456",
        "indices": [
          274,
          297
        ],
        "media_key": "3_2092900819368595456",
        "media_results": {
          "result": {
            "media_key": "3_2092900819368595456"
          }
        },
        "media_url_https": "https://pbs.twimg.com/media/HQt6NOCW0AAmda0.jpg",
        "original_info": {
          "focus_rects": [
            {
              "h": 722,
              "w": 1290,
              "x": 0,
              "y": 0
            },
            {
              "h": 1290,
              "w": 1290,
              "x": 0,
              "y": 0
            },
            {
              "h": 1471,
              "w": 1290,
              "x": 0,
              "y": 0
            },
            {
              "h": 1842,
              "w": 921,
              "x": 0,
              "y": 0
            },
            {
              "h": 1842,
              "w": 1290,
              "x": 0,
              "y": 0
            }
          ],
          "height": 1842,
          "width": 1290
        },
        "sizes": {
          "large": {
            "h": 1842,
            "resize": "fit",
            "w": 1290
          },
          "medium": {
            "h": 1200,
            "resize": "fit",
            "w": 840
          },
          "small": {
            "h": 680,
            "resize": "fit",
            "w": 476
          },
          "thumb": {
            "h": 150,
            "resize": "crop",
            "w": 150
          }
        },
        "type": "photo",
        "url": "https://t.co/vE4jOrvrq1"
      }
    ]
  },
  "card": null,
  "place": {},
  "entities": {
    "hashtags": [],
    "smarttags": [],
    "symbols": [],
    "urls": [
      {
        "display_url": "minjang10.github.io/demystifying-r…",
        "expanded_url": "https://minjang10.github.io/demystifying-rl-finetuning-web/",
        "indices": [
          625,
          648
        ],
        "url": "https://t.co/iPL9jqlxEL"
      },
      {
        "display_url": "github.com/sankarh-1/demy…",
        "expanded_url": "https://github.com/sankarh-1/demystifying-rl-finetuning",
        "indices": [
          655,
          678
        ],
        "url": "https://t.co/K9MtdC28Sl"
      },
      {
        "display_url": "arxiv.org/abs/2608.24949",
        "expanded_url": "https://arxiv.org/abs/2608.24949",
        "indices": [
          685,
          708
        ],
        "url": "https://t.co/RQfH9TqfLy"
      }
    ],
    "user_mentions": []
  },
  "quoted_tweet": null,
  "retweeted_tweet": null,
  "article": null
}