🐦 Twitter Post Details

Viewing enriched Twitter post

@nlpxuhui

Does LLM really need to be a helpful assistant all the time? No. If you want to simulate people, “perfectly helpful” could be the wrong objective. Meet OdysSim, a journey toward LLMs beyond assistants, as behavioral foundation models (10B tokens of real human behavior; 23 sim benchmarks, finally in one place. new open models: outperform or on par with GPT-5.5, Gemini 3.1, or Claude Opus 4.7 in many behavior-sim dimensions). Human behavior simulation is becoming essential. Agent evaluation needs realistic users before real users show up. Medical and classroom training need realistic patients and students. Social science needs synthetic participants at scale. But real people are not ideal assistants. Real patients panic or ignore good advice. Real students misunderstand. Real customers are vague, picky, impatient, or simply leave. Human behavior is messy, diverse, and often imperfect. Frontier LLMs are getting better at math, code, and long-horizon tasks. They are NOT getting better at simulating human behavior. If anything, they drift the other way: more assistant-ish, more homogeneous, fewer of the errors and quirks real humans show. This is no accident. The whole pipeline is built for helpfulness and task success, not behavioral realism. And you can't prompt your way out of that. So we rethink the recipe from scratch and release: 🧠 The OdysSim corpus: 21.4M real human interactions (~10B tokens) from 62 sources, every conversation retrofitted with social grounding (who is talking, and why) 📏 SOUL-Index: 23 human-behavior benchmarks unified into one suite across 5 axes 🤖 OSim-8B: open weights; tops more SOUL-Index benchmarks than any frontier model, acts more like a real user than any of them on τ-bench (nearly matching real humans in the reaction dimension), and writes far more human-like text along the way.

📊 Media Metadata

{
  "media": [
    {
      "url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2065111971700335030/media_0.mp4",
      "media_url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2065111971700335030/media_0.mp4",
      "type": "video",
      "filename": "media_0.mp4"
    }
  ],
  "processed_at": "2026-06-12T04:18:31.540156",
  "pipeline_version": "2.0"
}

🔧 Raw API Response

{
  "type": "tweet",
  "id": "2065111971700335030",
  "url": "https://x.com/nlpxuhui/status/2065111971700335030",
  "twitterUrl": "https://twitter.com/nlpxuhui/status/2065111971700335030",
  "text": "Does LLM really need to be a helpful assistant all the time?\n\nNo. If you want to simulate people, “perfectly helpful” could be the wrong objective.\n\nMeet OdysSim, a journey toward LLMs beyond assistants, as behavioral foundation models (10B tokens of real human behavior; 23 sim benchmarks, finally in one place. new open models: outperform or on par with GPT-5.5, Gemini 3.1, or Claude Opus 4.7 in many behavior-sim dimensions).\n\nHuman behavior simulation is becoming essential.\n\nAgent evaluation needs realistic users before real users show up. Medical and classroom training need realistic patients and students. Social science needs synthetic participants at scale.\n\nBut real people are not ideal assistants.\n\nReal patients panic or ignore good advice. Real students misunderstand. Real customers are vague, picky, impatient, or simply leave. Human behavior is messy, diverse, and often imperfect.\n\nFrontier LLMs are getting better at math, code, and long-horizon tasks. They are NOT getting better at simulating human behavior. If anything, they drift the other way: more assistant-ish, more homogeneous, fewer of the errors and quirks real humans show.\n\nThis is no accident. The whole pipeline is built for helpfulness and task success, not behavioral realism.\nAnd you can't prompt your way out of that. \n\nSo we rethink the recipe from scratch and release:\n\n🧠 The OdysSim corpus: 21.4M real human interactions (~10B tokens) from 62 sources, every conversation retrofitted with social grounding (who is talking, and why)\n📏 SOUL-Index: 23 human-behavior benchmarks unified into one suite across 5 axes\n🤖 OSim-8B: open weights; tops more SOUL-Index benchmarks than any frontier model, acts more like a real user than any of them on τ-bench (nearly matching real humans in the reaction dimension), and writes far more human-like text along the way.",
  "source": "Twitter for iPhone",
  "retweetCount": 31,
  "replyCount": 5,
  "likeCount": 163,
  "quoteCount": 3,
  "viewCount": 46933,
  "createdAt": "Thu Jun 11 16:40:36 +0000 2026",
  "lang": "en",
  "bookmarkCount": 130,
  "isReply": false,
  "inReplyToId": null,
  "conversationId": "2065111971700335030",
  "displayTextRange": [
    0,
    278
  ],
  "inReplyToUserId": null,
  "inReplyToUsername": null,
  "author": {
    "type": "user",
    "userName": "nlpxuhui",
    "url": "https://x.com/nlpxuhui",
    "twitterUrl": "https://twitter.com/nlpxuhui",
    "id": "1177300366620184576",
    "name": "Xuhui Zhou",
    "isVerified": false,
    "isBlueVerified": true,
    "verifiedType": null,
    "profilePicture": "https://pbs.twimg.com/profile_images/1584732418799386624/7Hvg39ab_normal.jpg",
    "coverPicture": "https://pbs.twimg.com/profile_banners/1177300366620184576/1666665730",
    "description": "",
    "location": "",
    "followers": 2008,
    "following": 810,
    "status": "",
    "canDm": false,
    "canMediaTag": true,
    "createdAt": "Thu Sep 26 19:14:34 +0000 2019",
    "entities": {
      "description": {
        "urls": []
      },
      "url": {}
    },
    "fastFollowersCount": 0,
    "favouritesCount": 4491,
    "hasCustomTimelines": true,
    "isTranslator": false,
    "mediaCount": 48,
    "statusesCount": 1804,
    "withheldInCountries": [],
    "affiliatesHighlightedLabel": {},
    "possiblySensitive": false,
    "pinnedTweetIds": [
      "2065111971700335030"
    ],
    "profile_bio": {
      "description": "PhD student @LTIatCMU. Previously, @openhandsdev, @allen_ai, @UWNLP, @Apple, @UCBerkeley; Social Intelligence in language +X.",
      "entities": {
        "description": {
          "user_mentions": [
            {
              "id_str": "",
              "indices": [
                12,
                21
              ],
              "name": "",
              "screen_name": "LTIatCMU"
            },
            {
              "id_str": "",
              "indices": [
                35,
                48
              ],
              "name": "",
              "screen_name": "openhandsdev"
            },
            {
              "id_str": "",
              "indices": [
                50,
                59
              ],
              "name": "",
              "screen_name": "allen_ai"
            },
            {
              "id_str": "",
              "indices": [
                61,
                67
              ],
              "name": "",
              "screen_name": "UWNLP"
            },
            {
              "id_str": "",
              "indices": [
                69,
                75
              ],
              "name": "",
              "screen_name": "Apple"
            },
            {
              "id_str": "",
              "indices": [
                77,
                88
              ],
              "name": "",
              "screen_name": "UCBerkeley"
            }
          ]
        },
        "url": {
          "urls": [
            {
              "display_url": "xuhuiz.com",
              "expanded_url": "https://xuhuiz.com/",
              "indices": [
                0,
                23
              ],
              "url": "https://t.co/7cccW6RuFQ"
            }
          ]
        }
      }
    },
    "isAutomated": false,
    "automatedBy": null
  },
  "extendedEntities": {
    "media": [
      {
        "additional_media_info": {
          "monetizable": true
        },
        "allow_download_status": {
          "allow_download": true
        },
        "display_url": "pic.twitter.com/IgI2PIJyKU",
        "expanded_url": "https://twitter.com/nlpxuhui/status/2065111971700335030/video/1",
        "ext_master_playlist_only": [],
        "ext_media_availability": {
          "status": "Available"
        },
        "ext_playlists": [],
        "id_str": "2065107899685462016",
        "indices": [
          279,
          302
        ],
        "media_key": "13_2065107899685462016",
        "media_results": {
          "id": "QXBpTWVkaWFSZXN1bHRzOgwABAoAARyovLF/muAAAAA=",
          "result": {
            "__typename": "ApiMedia",
            "id": "QXBpTWVkaWE6DAAECgABHKi8sX+a4AAAAA==",
            "media_key": "13_2065107899685462016"
          }
        },
        "media_url_https": "https://pbs.twimg.com/amplify_video_thumb/2065107899685462016/img/X7G4267NqUJsNAU4.jpg",
        "original_info": {
          "focus_rects": [],
          "height": 1410,
          "width": 1786
        },
        "sizes": {
          "large": {
            "h": 1410,
            "w": 1786
          }
        },
        "type": "video",
        "url": "https://t.co/IgI2PIJyKU",
        "video_info": {
          "aspect_ratio": [
            19,
            15
          ],
          "duration_millis": 17480,
          "variants": [
            {
              "content_type": "application/x-mpegURL",
              "url": "https://video.twimg.com/amplify_video/2065107899685462016/pl/G_AqCA4UOZSfIlPB.m3u8?tag=27"
            },
            {
              "bitrate": 256000,
              "content_type": "video/mp4",
              "url": "https://video.twimg.com/amplify_video/2065107899685462016/vid/avc1/342x270/qoe3VGvdujSm8HUe.mp4?tag=27"
            },
            {
              "bitrate": 832000,
              "content_type": "video/mp4",
              "url": "https://video.twimg.com/amplify_video/2065107899685462016/vid/avc1/456x360/nlwTAnqNyf80Rhog.mp4?tag=27"
            },
            {
              "bitrate": 2176000,
              "content_type": "video/mp4",
              "url": "https://video.twimg.com/amplify_video/2065107899685462016/vid/avc1/912x720/po6AeBVXPrrzINUU.mp4?tag=27"
            },
            {
              "bitrate": 10368000,
              "content_type": "video/mp4",
              "url": "https://video.twimg.com/amplify_video/2065107899685462016/vid/avc1/1786x1410/2GeGpMwZaQntSjW8.mp4?tag=27"
            }
          ]
        }
      }
    ]
  },
  "card": null,
  "place": {},
  "entities": {
    "hashtags": [],
    "symbols": [],
    "timestamps": [],
    "urls": [],
    "user_mentions": []
  },
  "quoted_tweet": null,
  "retweeted_tweet": null,
  "isLimitedReply": false,
  "communityInfo": null,
  "article": null
}