Viewing enriched Twitter post
Improved Techniques for Training Consistency Models abs: https://t.co/cOQEWAr8MT Consistency model training from scratch now beats all diffusion distillation approaches in both 1-step & 2-step generation. Comparable to many leading generative models, including diffusion models. This is possible with changes to loss weighting, noise schedule, metric in consistency loss, removing EMA, adding dropout, etc.
{
"media": [
{
"id": "",
"type": "photo",
"url": null,
"media_url": "https://pbs.twimg.com/media/F8XVaTyasAA0s2m.png",
"media_url_https": null,
"display_url": null,
"expanded_url": null
},
{
"id": "",
"type": "photo",
"url": null,
"media_url": "https://pbs.twimg.com/media/F8XVgxpawAAU0wq.jpg",
"media_url_https": null,
"display_url": null,
"expanded_url": null
},
{
"id": "",
"type": "photo",
"url": null,
"media_url": "https://pbs.twimg.com/media/F8XVnRcaIAATNMe.png",
"media_url_https": null,
"display_url": null,
"expanded_url": null
}
],
"nlp": {
"sentiment": "positive",
"processed_at": "2025-08-06T12:46:02.675722"
},
"original_structure": "had_media_only"
}{
"user": {
"created_at": "2011-12-20T03:45:50.000Z",
"default_profile_image": false,
"description": "PhD at 19 |\nFounder and CEO at @MedARC_AI |\nResearch Director at @StabilityAI | \n@kaggle Notebooks GM |\nBiomed. engineer @ 14 |\nTEDx talk➡https://t.co/DwMkst4bnG",
"fast_followers_count": 0,
"favourites_count": 60004,
"followers_count": 45439,
"friends_count": 995,
"has_custom_timelines": true,
"is_translator": false,
"listed_count": 703,
"location": "",
"media_count": 1203,
"name": "Tanishq Mathew Abraham, PhD",
"normal_followers_count": 45439,
"possibly_sensitive": false,
"profile_banner_url": "https://pbs.twimg.com/profile_banners/441465751/1675968078",
"profile_image_url_https": "https://pbs.twimg.com/profile_images/1553508977735962624/nnlSwBmu_normal.jpg",
"screen_name": "iScienceLuvr",
"statuses_count": 12087,
"translator_type": "none",
"url": "https://t.co/nNzCz2VVd1",
"verified": false,
"withheld_in_countries": [],
"id_str": "441465751"
},
"id": "1713010172824900061",
"conversation_id": "1713010172824900061",
"full_text": "Improved Techniques for Training Consistency Models\n\nabs: https://t.co/cOQEWAr8MT\n\nConsistency model training from scratch now beats all diffusion distillation approaches in both 1-step & 2-step generation. Comparable to many leading generative models, including diffusion models.\n\nThis is possible with changes to loss weighting, noise schedule, metric in consistency loss, removing EMA, adding dropout, etc.",
"reply_count": 2,
"retweet_count": 22,
"favorite_count": 112,
"hashtags": [],
"symbols": [],
"user_mentions": [],
"urls": [
{
"url": "https://t.co/4lghBp4Fhl",
"expanded_url": "https://openreview.net/forum?id=WNzy9bRDvG",
"display_url": "openreview.net/forum?id=WNzy9…"
}
],
"media": [
{
"media_url": "https://pbs.twimg.com/media/F8XVaTyasAA0s2m.png",
"type": "photo"
},
{
"media_url": "https://pbs.twimg.com/media/F8XVgxpawAAU0wq.jpg",
"type": "photo"
},
{
"media_url": "https://pbs.twimg.com/media/F8XVnRcaIAATNMe.png",
"type": "photo"
}
],
"url": "https://twitter.com/iScienceLuvr/status/1713010172824900061",
"created_at": "2023-10-14T01:53:46.000Z",
"#sort_index": "1713010172824900061",
"view_count": 28187,
"quote_count": 1,
"is_quote_tweet": false,
"is_retweet": false,
"is_pinned": false,
"is_truncated": true,
"startUrl": "https://twitter.com/iscienceluvr/status/1713010172824900061"
}