🐦 Twitter Post Details

Viewing enriched Twitter post

@omarsar0

NEW paper from Microsoft Every agent benchmark has the same hidden problem: how do you know the agent actually succeeded? Microsoft researchers introduce the Universal Verifier, which discusses lessons learned from building best-in-class verifiers for web tasks. It's built on four principles: non-overlapping rubrics, separate process vs. outcome rewards, distinguishing controllable from uncontrollable failures, and divide-and-conquer context management across full screenshot trajectories. It reduces false positive rates to near zero, down from 45%+ (WebVoyager) and 22%+ (WebJudge). Without reliable verifiers, both benchmarks and training data are corrupted. One interesting finding is that an auto-research agent reached 70% of expert verifier quality in 5% of the time, but couldn't discover the structural design decisions that drove the biggest gains. Human expertise and automated optimization play complementary roles. Paper: https://t.co/fWhG9I8vPP Learn to build effective AI agents in our academy: https://t.co/1e8RZKs4uX

Media 1
View on Twitter

📊 Media Metadata

{
  "media": [
    {
      "type": "photo",
      "url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2042249194409501054/media_0.jpg",
      "filename": "media_0.jpg"
    }
  ],
  "processed_at": "2026-04-09T14:37:03.808124",
  "pipeline_version": "2.0"
}

🔧 Raw API Response

{
  "type": "tweet",
  "id": "2042249194409501054",
  "url": "https://x.com/omarsar0/status/2042249194409501054",
  "twitterUrl": "https://twitter.com/omarsar0/status/2042249194409501054",
  "text": "NEW paper from Microsoft\n\nEvery agent benchmark has the same hidden problem: how do you know the agent actually succeeded?\n\nMicrosoft researchers introduce the Universal Verifier, which discusses lessons learned from building best-in-class verifiers for web tasks.\n\nIt's built on four principles: non-overlapping rubrics, separate process vs. outcome rewards, distinguishing controllable from uncontrollable failures, and divide-and-conquer context management across full screenshot trajectories.\n\nIt reduces false positive rates to near zero, down from 45%+ (WebVoyager) and 22%+ (WebJudge).\n\nWithout reliable verifiers, both benchmarks and training data are corrupted.\n\nOne interesting finding is that an auto-research agent reached 70% of expert verifier quality in 5% of the time, but couldn't discover the structural design decisions that drove the biggest gains. Human expertise and automated optimization play complementary roles.\n\nPaper: https://t.co/fWhG9I8vPP\n\nLearn to build effective AI agents in our academy: https://t.co/1e8RZKs4uX",
  "source": "Twitter for iPhone",
  "retweetCount": 0,
  "replyCount": 0,
  "likeCount": 5,
  "quoteCount": 0,
  "viewCount": 229,
  "createdAt": "Thu Apr 09 14:32:05 +0000 2026",
  "lang": "en",
  "bookmarkCount": 6,
  "isReply": false,
  "inReplyToId": null,
  "conversationId": "2042249194409501054",
  "displayTextRange": [
    0,
    279
  ],
  "inReplyToUserId": null,
  "inReplyToUsername": null,
  "author": {
    "type": "user",
    "userName": "omarsar0",
    "url": "https://x.com/omarsar0",
    "twitterUrl": "https://twitter.com/omarsar0",
    "id": "3448284313",
    "name": "elvis",
    "isVerified": false,
    "isBlueVerified": true,
    "verifiedType": null,
    "profilePicture": "https://pbs.twimg.com/profile_images/939313677647282181/vZjFWtAn_normal.jpg",
    "coverPicture": "https://pbs.twimg.com/profile_banners/3448284313/1565974901",
    "description": "",
    "location": "DAIR.AI Academy",
    "followers": 297651,
    "following": 801,
    "status": "",
    "canDm": true,
    "canMediaTag": true,
    "createdAt": "Fri Sep 04 12:59:26 +0000 2015",
    "entities": {
      "description": {
        "urls": []
      },
      "url": {}
    },
    "fastFollowersCount": 0,
    "favouritesCount": 35331,
    "hasCustomTimelines": true,
    "isTranslator": true,
    "mediaCount": 4587,
    "statusesCount": 17598,
    "withheldInCountries": [],
    "affiliatesHighlightedLabel": {},
    "possiblySensitive": false,
    "pinnedTweetIds": [
      "2041901555280945558"
    ],
    "profile_bio": {
      "description": "Building @dair_ai • Prev: Meta AI, Elastic, PhD • New AI learning portal: https://t.co/1e8RZKs4uX",
      "entities": {
        "description": {
          "hashtags": [],
          "symbols": [],
          "urls": [
            {
              "display_url": "academy.dair.ai",
              "expanded_url": "https://academy.dair.ai/",
              "indices": [
                74,
                97
              ],
              "url": "https://t.co/1e8RZKs4uX"
            }
          ],
          "user_mentions": [
            {
              "id_str": "0",
              "indices": [
                9,
                17
              ],
              "name": "",
              "screen_name": "dair_ai"
            }
          ]
        },
        "url": {
          "urls": [
            {
              "display_url": "dair.ai",
              "expanded_url": "https://www.dair.ai/",
              "indices": [
                0,
                23
              ],
              "url": "https://t.co/XQto5ypSIk"
            }
          ]
        }
      }
    },
    "isAutomated": false,
    "automatedBy": null
  },
  "extendedEntities": {
    "media": [
      {
        "display_url": "pic.twitter.com/EBxg3et9zl",
        "expanded_url": "https://twitter.com/omarsar0/status/2042249194409501054/photo/1",
        "ext_media_availability": {
          "status": "Available"
        },
        "features": {
          "large": {
            "faces": []
          },
          "orig": {
            "faces": []
          }
        },
        "id_str": "2042249191389536256",
        "indices": [
          280,
          303
        ],
        "media_key": "3_2042249191389536256",
        "media_results": {
          "id": "QXBpTWVkaWFSZXN1bHRzOgwAAQoAARxXhtG7miAACgACHFeG0m+bIX4AAA==",
          "result": {
            "__typename": "ApiMedia",
            "id": "QXBpTWVkaWE6DAABCgABHFeG0buaIAAKAAIcV4bSb5shfgAA",
            "media_key": "3_2042249191389536256"
          }
        },
        "media_url_https": "https://pbs.twimg.com/media/HFeG0buaIAA1sls.jpg",
        "original_info": {
          "focus_rects": [
            {
              "h": 783,
              "w": 1398,
              "x": 0,
              "y": 0
            },
            {
              "h": 1398,
              "w": 1398,
              "x": 0,
              "y": 0
            },
            {
              "h": 1568,
              "w": 1375,
              "x": 23,
              "y": 0
            },
            {
              "h": 1568,
              "w": 784,
              "x": 352,
              "y": 0
            },
            {
              "h": 1568,
              "w": 1398,
              "x": 0,
              "y": 0
            }
          ],
          "height": 1568,
          "width": 1398
        },
        "sizes": {
          "large": {
            "h": 1568,
            "w": 1398
          }
        },
        "type": "photo",
        "url": "https://t.co/EBxg3et9zl"
      }
    ]
  },
  "card": null,
  "place": {},
  "entities": {
    "hashtags": [],
    "symbols": [],
    "urls": [
      {
        "display_url": "arxiv.org/abs/2604.06240",
        "expanded_url": "https://arxiv.org/abs/2604.06240",
        "indices": [
          946,
          969
        ],
        "url": "https://t.co/fWhG9I8vPP"
      },
      {
        "display_url": "academy.dair.ai",
        "expanded_url": "https://academy.dair.ai/",
        "indices": [
          1022,
          1045
        ],
        "url": "https://t.co/1e8RZKs4uX"
      }
    ],
    "user_mentions": []
  },
  "quoted_tweet": null,
  "retweeted_tweet": null,
  "isLimitedReply": false,
  "communityInfo": null,
  "article": null
}