🐦 Twitter Post Details

Viewing enriched Twitter post

@dair_ai

Great paper from NVIDIA sharing an open midtraining recipe for agentic coding. NVIDIA kept SFT identical and added one midtraining stage. SWE-bench Multilingual went from 55.8 to 62.8 on Nemotron 3.5 Lightning. The stage sits between pretraining and post-training. It trains at 256K context on 500B tokens of natural software-engineering data, such as packed repositories, pull requests, commit histories and build logs, plus a small share of agent traces. On the same model, SWE-bench Verified rose from 60.7 to 64.5 and SWE-bench Pro from 32.0 to 35.5. On the larger Nemotron 3.5 Super, DeepSWE rose from 11.7 to 15.7. They also explain how they evaluated the recipe. Base-model proxy metrics did not predict post-training results, so every data choice was scored with short fixed SFT runs. If you train coding agents, the recipe and its ablations are published. https://t.co/SJ7JHdd1HY Chat with Paper: https://t.co/eCK8iUNIi0

Media 1
View on Twitter

📊 Media Metadata

{
  "media": [
    {
      "type": "photo",
      "url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2111452428294144180/media_0.jpg",
      "filename": "media_0.jpg"
    }
  ],
  "processed_at": "2026-10-17T15:02:28.705732",
  "pipeline_version": "2.0"
}

🔧 Raw API Response

{
  "type": "tweet",
  "id": "2111452428294144180",
  "url": "https://x.com/dair_ai/status/2111452428294144180",
  "twitterUrl": "https://twitter.com/dair_ai/status/2111452428294144180",
  "text": "Great paper from NVIDIA sharing an open midtraining recipe for agentic coding.\n\nNVIDIA kept SFT identical and added one midtraining stage. SWE-bench Multilingual went from 55.8 to 62.8 on Nemotron 3.5 Lightning.\n\nThe stage sits between pretraining and post-training. It trains at 256K context on 500B tokens of natural software-engineering data, such as packed repositories, pull requests, commit histories and build logs, plus a small share of agent traces.\n\nOn the same model, SWE-bench Verified rose from 60.7 to 64.5 and SWE-bench Pro from 32.0 to 35.5. On the larger Nemotron 3.5 Super, DeepSWE rose from 11.7 to 15.7.\n\nThey also explain how they evaluated the recipe. Base-model proxy metrics did not predict post-training results, so every data choice was scored with short fixed SFT runs.\n\nIf you train coding agents, the recipe and its ablations are published.\n\nhttps://t.co/SJ7JHdd1HY\n\nChat with Paper: https://t.co/eCK8iUNIi0",
  "source": "Twitter for iPhone",
  "retweetCount": 5,
  "replyCount": 8,
  "likeCount": 33,
  "quoteCount": 0,
  "viewCount": 2626,
  "createdAt": "Sat Oct 17 13:41:02 +0000 2026",
  "lang": "en",
  "bookmarkCount": 17,
  "isReply": false,
  "inReplyToId": null,
  "conversationId": "2111452428294144180",
  "displayTextRange": [
    0,
    279
  ],
  "inReplyToUserId": null,
  "inReplyToUsername": null,
  "author": {
    "type": "user",
    "userName": "dair_ai",
    "url": "https://x.com/dair_ai",
    "twitterUrl": "https://twitter.com/dair_ai",
    "id": "889050642903293953",
    "name": "DAIR.AI",
    "isVerified": false,
    "isBlueVerified": true,
    "verifiedType": null,
    "profilePicture": "https://pbs.twimg.com/profile_images/1643277398522187778/31dedbLo_normal.jpg",
    "coverPicture": "https://pbs.twimg.com/profile_banners/889050642903293953/1773242460",
    "description": "",
    "location": "",
    "followers": 134763,
    "following": 1,
    "status": "",
    "canDm": true,
    "canMediaTag": true,
    "createdAt": "Sun Jul 23 09:12:45 +0000 2017",
    "entities": {
      "description": {
        "urls": []
      },
      "url": {}
    },
    "fastFollowersCount": 0,
    "favouritesCount": 5628,
    "hasCustomTimelines": true,
    "isTranslator": false,
    "mediaCount": 440,
    "statusesCount": 4000,
    "withheldInCountries": [],
    "affiliatesHighlightedLabel": {},
    "possiblySensitive": false,
    "pinnedTweetIds": [
      "2111155277886316847"
    ],
    "profile_bio": {
      "description": "Democratizing AI research, education, and technologies. Learn about AI Agents for FREE at https://t.co/HHXg8rryu4",
      "entities": {
        "description": {
          "urls": [
            {
              "display_url": "academy.dair.ai/courses/elemen…",
              "expanded_url": "https://academy.dair.ai/courses/elements-of-ai-agents",
              "indices": [
                90,
                113
              ],
              "url": "https://t.co/HHXg8rryu4"
            }
          ]
        },
        "url": {
          "urls": [
            {
              "display_url": "dair.ai",
              "expanded_url": "https://www.dair.ai/",
              "indices": [
                0,
                23
              ],
              "url": "https://t.co/lkqPZtMU5s"
            }
          ]
        }
      }
    },
    "isAutomated": false,
    "automatedBy": null
  },
  "extendedEntities": {
    "media": [
      {
        "display_url": "pic.x.com/EMhRzXqgXd",
        "expanded_url": "https://x.com/dair_ai/status/2111452428294144180/photo/1",
        "ext_master_playlist_only": [],
        "ext_media_availability": {
          "status": "Available"
        },
        "ext_playlists": [],
        "features": {
          "large": {
            "faces": []
          },
          "orig": {
            "faces": []
          }
        },
        "id_str": "2111452424380776448",
        "indices": [
          280,
          303
        ],
        "media_key": "3_2111452424380776448",
        "media_results": {
          "id": "QXBpTWVkaWFSZXN1bHRzOgwAAQoAAR1NYssDWoAACgACHU1iy+ybwLQAAA==",
          "result": {
            "__typename": "ApiMedia",
            "id": "QXBpTWVkaWE6DAABCgABHU1iywNagAAKAAIdTWLL7JvAtAAA",
            "media_key": "3_2111452424380776448"
          }
        },
        "media_url_https": "https://pbs.twimg.com/media/HU1iywNagAAogTH.jpg",
        "original_info": {
          "focus_rects": [
            {
              "h": 913,
              "w": 1630,
              "x": 0,
              "y": 0
            },
            {
              "h": 1630,
              "w": 1630,
              "x": 0,
              "y": 0
            },
            {
              "h": 1858,
              "w": 1630,
              "x": 0,
              "y": 0
            },
            {
              "h": 1878,
              "w": 939,
              "x": 346,
              "y": 0
            },
            {
              "h": 1878,
              "w": 1630,
              "x": 0,
              "y": 0
            }
          ],
          "height": 1878,
          "width": 1630
        },
        "sizes": {
          "large": {
            "h": 1878,
            "w": 1630
          }
        },
        "type": "photo",
        "url": "https://t.co/EMhRzXqgXd"
      }
    ]
  },
  "card": null,
  "place": {},
  "entities": {
    "hashtags": [],
    "symbols": [],
    "urls": [
      {
        "display_url": "arxiv.org/abs/2610.22019",
        "expanded_url": "https://arxiv.org/abs/2610.22019",
        "indices": [
          871,
          894
        ],
        "url": "https://t.co/SJ7JHdd1HY"
      },
      {
        "display_url": "academy.dair.ai/papers/priming…",
        "expanded_url": "https://academy.dair.ai/papers/priming-base-models-for-agentic-post-training-an-open-midtraining-recipe-2610.22019",
        "indices": [
          913,
          936
        ],
        "url": "https://t.co/eCK8iUNIi0"
      }
    ],
    "user_mentions": []
  },
  "quoted_tweet": null,
  "retweeted_tweet": null,
  "isLimitedReply": false,
  "communityInfo": null,
  "article": null
}