🐦 Twitter Post Details

Viewing enriched Twitter post

@jerryjliu0

We’re open sourcing the first document OCR benchmark for the agentic era, ParseBench. Document parsing is the foundation of every AI agent that works with real-world files. ParseBench is a benchmark that measures parsing quality specifically for agent knowledge work: ✅ It optimizes for semantic correctness (instead of exact similarity) ✅ It has the most comprehensive distribution of real-world enterprise documents It contains ~2,000 human-verified enterprise document pages with 167,000+ test rules across five dimensions that matter most: tables, charts, content faithfulness, semantic formatting, and visual grounding. We benchmarked 14 known document parsers on ParseBench, from frontier/OSS VLMs to specialized parsers to LlamaParse. Here are some of our findings: 💡 Increasing compute budget yields diminishing returns - Gemini/gpt-5-mini/haiku gain 3-5 points from minimal to high thinking, at 4x the cost. 💡 Charts are the most polarizing dimension for evaluation. Most specialized parsers score below 6%, while some VLM-based parsers do a bit better. 💡 VLMs are great at visual understanding but terrible at layout extraction. GPT-5-mini/haiku score below 10% on our visual grounding task, all specialized parsers do much better. 💡 No method crushes all 5 dimensions at once, but LlamaParse achieves the highest overall score at 84.9%, and is the leader in 4 out of the 5 dimensions. This is by far the deepest technical work that we’ve published as a company. I would encourage you to start with our blog and explore our links to Hugging Face to GitHub. All the details are in our full 35-page (!!) ArXiv whitepaper. 🌐: Blog: https://t.co/57OHkx0pQW 📄 Paper: https://t.co/Ho2oH2xEAM 💻 Code: https://t.co/6P7UxqOZYA 📊 Dataset: https://t.co/YguIXWm41j 🎥 YouTube: https://t.co/6Fh1Nsk9ei

Media 2
Media 3
Media 4

📊 Media Metadata

{
  "media": [
    {
      "url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2043721536922955918/media_0.mp4",
      "media_url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2043721536922955918/media_0.mp4",
      "type": "video",
      "filename": "media_0.mp4"
    },
    {
      "url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2043721536922955918/media_1.jpg",
      "media_url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2043721536922955918/media_1.jpg",
      "type": "photo",
      "filename": "media_1.jpg"
    },
    {
      "url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2043721536922955918/media_3.jpg",
      "media_url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2043721536922955918/media_3.jpg",
      "type": "photo",
      "filename": "media_3.jpg"
    },
    {
      "url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2043721536922955918/media_4.jpg",
      "media_url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2043721536922955918/media_4.jpg",
      "type": "photo",
      "filename": "media_4.jpg"
    }
  ],
  "processed_at": "2026-04-13T16:34:13.784850",
  "pipeline_version": "2.0"
}

🔧 Raw API Response

{
  "type": "tweet",
  "id": "2043721536922955918",
  "url": "https://x.com/jerryjliu0/status/2043721536922955918",
  "twitterUrl": "https://twitter.com/jerryjliu0/status/2043721536922955918",
  "text": "We’re open sourcing the first document OCR benchmark for the agentic era, ParseBench.\n\nDocument parsing is the foundation of every AI agent that works with real-world files. ParseBench is a benchmark that measures parsing quality specifically for agent knowledge work:\n✅ It optimizes for semantic correctness (instead of exact similarity)\n✅ It has the most comprehensive distribution of real-world enterprise documents\n\nIt contains ~2,000 human-verified enterprise document pages with 167,000+ test rules across five dimensions that matter most: tables, charts, content faithfulness, semantic formatting, and visual grounding.\n\nWe benchmarked 14 known document parsers on ParseBench, from frontier/OSS VLMs to specialized parsers to LlamaParse. Here are some of our findings:\n\n💡 Increasing compute budget yields diminishing returns - Gemini/gpt-5-mini/haiku gain 3-5 points from minimal to high thinking, at 4x the cost.\n\n💡 Charts are the most polarizing dimension for evaluation. Most specialized parsers score below 6%, while some VLM-based parsers do a bit better.\n\n💡 VLMs are great at visual understanding but terrible at layout extraction. GPT-5-mini/haiku score below 10% on our visual grounding task, all specialized parsers do much better.\n\n💡 No method crushes all 5 dimensions at once, but LlamaParse achieves the highest overall score at 84.9%, and is the leader in 4 out of the 5 dimensions.\n\nThis is by far the deepest technical work that we’ve published as a company. I would encourage you to start with our blog and explore our links to Hugging Face to GitHub. All the details are in our full 35-page (!!) ArXiv whitepaper.\n\n🌐: Blog: https://t.co/57OHkx0pQW\n\n📄 Paper: https://t.co/Ho2oH2xEAM\n\n💻 Code: https://t.co/6P7UxqOZYA\n\n📊 Dataset: https://t.co/YguIXWm41j\n\n🎥 YouTube: https://t.co/6Fh1Nsk9ei",
  "source": "Twitter for iPhone",
  "retweetCount": 11,
  "replyCount": 0,
  "likeCount": 36,
  "quoteCount": 1,
  "viewCount": 2037,
  "createdAt": "Mon Apr 13 16:02:39 +0000 2026",
  "lang": "en",
  "bookmarkCount": 26,
  "isReply": false,
  "inReplyToId": null,
  "conversationId": "2043721536922955918",
  "displayTextRange": [
    0,
    273
  ],
  "inReplyToUserId": null,
  "inReplyToUsername": null,
  "author": {
    "type": "user",
    "userName": "jerryjliu0",
    "url": "https://x.com/jerryjliu0",
    "twitterUrl": "https://twitter.com/jerryjliu0",
    "id": "369777416",
    "name": "Jerry Liu",
    "isVerified": false,
    "isBlueVerified": true,
    "verifiedType": null,
    "profilePicture": "https://pbs.twimg.com/profile_images/1283610285031460864/1Q4zYhtb_normal.jpg",
    "coverPicture": "",
    "description": "",
    "location": "",
    "followers": 72598,
    "following": 1470,
    "status": "",
    "canDm": true,
    "canMediaTag": true,
    "createdAt": "Wed Sep 07 22:54:31 +0000 2011",
    "entities": {
      "description": {
        "urls": []
      },
      "url": {}
    },
    "fastFollowersCount": 0,
    "favouritesCount": 8636,
    "hasCustomTimelines": true,
    "isTranslator": false,
    "mediaCount": 1480,
    "statusesCount": 6838,
    "withheldInCountries": [],
    "affiliatesHighlightedLabel": {},
    "possiblySensitive": false,
    "pinnedTweetIds": [],
    "profile_bio": {
      "description": "Parsing the world's hardest PDFs @llama_index. cofounder/CEO\n\nCareers: https://t.co/EUnMNmbCtx\nEnterprise: https://t.co/Ht5jwxSrQB",
      "entities": {
        "description": {
          "hashtags": [],
          "symbols": [],
          "urls": [
            {
              "display_url": "llamaindex.ai/careers",
              "expanded_url": "https://www.llamaindex.ai/careers",
              "indices": [
                71,
                94
              ],
              "url": "https://t.co/EUnMNmbCtx"
            },
            {
              "display_url": "llamaindex.ai/contact",
              "expanded_url": "https://www.llamaindex.ai/contact",
              "indices": [
                107,
                130
              ],
              "url": "https://t.co/Ht5jwxSrQB"
            }
          ],
          "user_mentions": [
            {
              "id_str": "0",
              "indices": [
                33,
                45
              ],
              "name": "",
              "screen_name": "llama_index"
            }
          ]
        },
        "url": {
          "urls": [
            {
              "display_url": "llamaindex.ai",
              "expanded_url": "https://www.llamaindex.ai/",
              "indices": [
                0,
                23
              ],
              "url": "https://t.co/YiIfjVlzb6"
            }
          ]
        }
      }
    },
    "isAutomated": false,
    "automatedBy": null
  },
  "extendedEntities": {
    "media": [
      {
        "additional_media_info": {
          "monetizable": true
        },
        "display_url": "pic.twitter.com/834eBWhrDk",
        "expanded_url": "https://twitter.com/jerryjliu0/status/2043721536922955918/video/1",
        "ext_media_availability": {
          "status": "Available"
        },
        "id_str": "2043719850007400448",
        "indices": [
          274,
          297
        ],
        "media_key": "13_2043719850007400448",
        "media_results": {
          "id": "QXBpTWVkaWFSZXN1bHRzOgwABAoAARxcwGAgGxAAAAA=",
          "result": {
            "__typename": "ApiMedia",
            "id": "QXBpTWVkaWE6DAAECgABHFzAYCAbEAAAAA==",
            "media_key": "13_2043719850007400448"
          }
        },
        "media_url_https": "https://pbs.twimg.com/amplify_video_thumb/2043719850007400448/img/RgxAds0ItR91Qg1I.jpg",
        "original_info": {
          "focus_rects": [],
          "height": 2160,
          "width": 3840
        },
        "sizes": {
          "large": {
            "h": 1152,
            "w": 2048
          }
        },
        "type": "video",
        "url": "https://t.co/834eBWhrDk",
        "video_info": {
          "aspect_ratio": [
            16,
            9
          ],
          "duration_millis": 68601,
          "variants": [
            {
              "content_type": "application/x-mpegURL",
              "url": "https://video.twimg.com/amplify_video/2043719850007400448/pl/IVEM6hG7F1k33i-f.m3u8?tag=21"
            },
            {
              "bitrate": 256000,
              "content_type": "video/mp4",
              "url": "https://video.twimg.com/amplify_video/2043719850007400448/vid/avc1/480x270/u5XXueRDViKNYkpM.mp4?tag=21"
            },
            {
              "bitrate": 832000,
              "content_type": "video/mp4",
              "url": "https://video.twimg.com/amplify_video/2043719850007400448/vid/avc1/640x360/mWYrAFjjnhB2tvhg.mp4?tag=21"
            },
            {
              "bitrate": 2176000,
              "content_type": "video/mp4",
              "url": "https://video.twimg.com/amplify_video/2043719850007400448/vid/avc1/1280x720/FWIOSN6eO5FeoDWH.mp4?tag=21"
            },
            {
              "bitrate": 10368000,
              "content_type": "video/mp4",
              "url": "https://video.twimg.com/amplify_video/2043719850007400448/vid/avc1/1920x1080/gGahZTF-NqaHT6ZT.mp4?tag=21"
            },
            {
              "bitrate": 25128000,
              "content_type": "video/mp4",
              "url": "https://video.twimg.com/amplify_video/2043719850007400448/vid/avc1/3840x2160/UT1gxueJcgoqfLKm.mp4?tag=21"
            }
          ]
        }
      }
    ]
  },
  "card": null,
  "place": {},
  "entities": {
    "hashtags": [],
    "symbols": [],
    "timestamps": [],
    "urls": [
      {
        "display_url": "llamaindex.ai/blog/parsebenc…",
        "expanded_url": "https://www.llamaindex.ai/blog/parsebench?utm_medium=socials&utm_source=xjl&utm_campaign=2026-apr-",
        "indices": [
          1648,
          1671
        ],
        "url": "https://t.co/57OHkx0pQW"
      },
      {
        "display_url": "arxiv.org/abs/2604.08538…",
        "expanded_url": "https://arxiv.org/abs/2604.08538?utm_medium=socials&utm_source=twitter&utm_campaign=2026-apr-",
        "indices": [
          1682,
          1705
        ],
        "url": "https://t.co/Ho2oH2xEAM"
      },
      {
        "display_url": "github.com/run-llama/Pars…",
        "expanded_url": "https://github.com/run-llama/ParseBench?utm_medium=socials&utm_source=twitter&utm_campaign=2026-apr-",
        "indices": [
          1715,
          1738
        ],
        "url": "https://t.co/6P7UxqOZYA"
      },
      {
        "display_url": "huggingface.co/datasets/llama…",
        "expanded_url": "https://huggingface.co/datasets/llamaindex/ParseBench?utm_medium=socials&utm_source=twitter&utm_campaign=2026-apr-",
        "indices": [
          1751,
          1774
        ],
        "url": "https://t.co/YguIXWm41j"
      },
      {
        "display_url": "youtube.com/watch?v=g5p7G-…",
        "expanded_url": "https://www.youtube.com/watch?v=g5p7G-Nw8pQ?utm_medium=socials&utm_source=twitter&utm_campaign=2026-apr-",
        "indices": [
          1787,
          1810
        ],
        "url": "https://t.co/6Fh1Nsk9ei"
      }
    ],
    "user_mentions": []
  },
  "quoted_tweet": null,
  "retweeted_tweet": null,
  "isLimitedReply": false,
  "communityInfo": null,
  "article": null
}