Viewing enriched Twitter post
Imagination Helps Visual Reasoning, But Not Yet in Latent Space Causal mediation analysis reveals latent visual reasoning in MLLMs fails: latent tokens ignore inputs and barely affect answers. CapImagine, a text-based alternative, teaches explicit imagination and significantly outperforms latent baselines.
{
"media": [
{
"url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2027539563573370956/media_0.jpg?",
"media_url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2027539563573370956/media_0.jpg?",
"type": "photo",
"filename": "media_0.jpg"
}
],
"processed_at": "2026-03-01T19:35:37.193543",
"pipeline_version": "2.0"
}{
"type": "tweet",
"id": "2027539563573370956",
"url": "https://x.com/HuggingPapers/status/2027539563573370956",
"twitterUrl": "https://twitter.com/HuggingPapers/status/2027539563573370956",
"text": "Imagination Helps Visual Reasoning, But Not Yet in Latent Space\n\nCausal mediation analysis reveals latent visual reasoning in MLLMs fails: latent tokens ignore inputs and barely affect answers. CapImagine, a text-based alternative, teaches explicit imagination and significantly outperforms latent baselines.",
"source": "Twitter for iPhone",
"retweetCount": 8,
"replyCount": 5,
"likeCount": 38,
"quoteCount": 0,
"viewCount": 8139,
"createdAt": "Sat Feb 28 00:21:16 +0000 2026",
"lang": "en",
"bookmarkCount": 20,
"isReply": false,
"inReplyToId": null,
"conversationId": "2027539563573370956",
"displayTextRange": [
0,
278
],
"inReplyToUserId": null,
"inReplyToUsername": null,
"author": {
"type": "user",
"userName": "HuggingPapers",
"url": "https://x.com/HuggingPapers",
"twitterUrl": "https://twitter.com/HuggingPapers",
"id": "1906617820122595328",
"name": "DailyPapers",
"isVerified": false,
"isBlueVerified": true,
"verifiedType": null,
"profilePicture": "https://pbs.twimg.com/profile_images/1906618607368273920/09C2Z22U_normal.jpg",
"coverPicture": "https://pbs.twimg.com/profile_banners/1906617820122595328/1743408239",
"description": "",
"location": "Anywhere",
"followers": 14290,
"following": 4,
"status": "",
"canDm": false,
"canMediaTag": true,
"createdAt": "Mon Mar 31 08:01:39 +0000 2025",
"entities": {
"description": {
"urls": []
},
"url": {}
},
"fastFollowersCount": 0,
"favouritesCount": 23,
"hasCustomTimelines": true,
"isTranslator": false,
"mediaCount": 1935,
"statusesCount": 4244,
"withheldInCountries": [],
"affiliatesHighlightedLabel": {},
"possiblySensitive": false,
"pinnedTweetIds": [],
"profile_bio": {
"description": "Tweeting interesting papers submitted at https://t.co/rXX8x0HzXV.\n\nSubmit your own at https://t.co/QhbJKXBd4Q, and link models/datasets/demos to it!",
"entities": {
"description": {
"hashtags": [],
"symbols": [],
"urls": [
{
"display_url": "huggingface.co/papers",
"expanded_url": "http://huggingface.co/papers",
"indices": [
41,
64
],
"url": "https://t.co/rXX8x0HzXV"
},
{
"display_url": "hf.co/papers/submit",
"expanded_url": "http://hf.co/papers/submit",
"indices": [
86,
109
],
"url": "https://t.co/QhbJKXBd4Q"
}
],
"user_mentions": []
},
"url": {
"urls": [
{
"display_url": "huggingface.co/docs/hub/en/mo…",
"expanded_url": "https://huggingface.co/docs/hub/en/model-release-checklist",
"indices": [
0,
23
],
"url": "https://t.co/rEWlTAkfrC"
}
]
}
}
},
"isAutomated": false,
"automatedBy": null
},
"extendedEntities": {
"media": [
{
"display_url": "pic.twitter.com/OBozJ3DXCF",
"expanded_url": "https://twitter.com/HuggingPapers/status/2027539563573370956/photo/1",
"ext_media_availability": {
"status": "Available"
},
"features": {
"large": {
"faces": [
{
"h": 63,
"w": 63,
"x": 1361,
"y": 75
},
{
"h": 142,
"w": 142,
"x": 1554,
"y": 675
}
]
},
"orig": {
"faces": [
{
"h": 63,
"w": 63,
"x": 1361,
"y": 75
},
{
"h": 142,
"w": 142,
"x": 1554,
"y": 675
}
]
}
},
"id_str": "2027539558334926850",
"indices": [
279,
302
],
"media_key": "3_2027539558334926850",
"media_results": {
"id": "QXBpTWVkaWFSZXN1bHRzOgwAAQoAARwjRHxmmxACCgACHCNEfZ7XYEwAAA==",
"result": {
"__typename": "ApiMedia",
"id": "QXBpTWVkaWE6DAABCgABHCNEfGabEAIKAAIcI0R9ntdgTAAA",
"media_key": "3_2027539558334926850"
}
},
"media_url_https": "https://pbs.twimg.com/media/HCNEfGabEAItqBx.jpg",
"original_info": {
"focus_rects": [
{
"h": 880,
"w": 1571,
"x": 0,
"y": 0
},
{
"h": 880,
"w": 880,
"x": 0,
"y": 0
},
{
"h": 880,
"w": 772,
"x": 0,
"y": 0
},
{
"h": 880,
"w": 440,
"x": 19,
"y": 0
},
{
"h": 880,
"w": 1920,
"x": 0,
"y": 0
}
],
"height": 880,
"width": 1920
},
"sizes": {
"large": {
"h": 880,
"w": 1920
}
},
"type": "photo",
"url": "https://t.co/OBozJ3DXCF"
}
]
},
"card": null,
"place": {},
"entities": {
"hashtags": [],
"symbols": [],
"urls": [],
"user_mentions": []
},
"quoted_tweet": null,
"retweeted_tweet": null,
"isLimitedReply": false,
"article": null
}