Viewing enriched Twitter post
gte-tiny not small enough for you? Consider bge-micro, an embeddings model with 1/4 the layers of bge-small. It's not SOTA—compressing a model this much does incur degradation—but it's still good enough to be competitive with all-MiniLM-L6-v2, with 1/2 the non-embedding params!
{
"media": [
{
"id": "",
"type": "photo",
"url": null,
"media_url": "https://pbs.twimg.com/media/F72n0vpb0AAA3Lf.png",
"media_url_https": null,
"display_url": null,
"expanded_url": null
}
],
"nlp": {
"processed_at": "2025-08-06T12:42:28.931018",
"sentiment": "neutral",
"topics": [
"Machine Learning",
"Deep Learning",
"Neural Networks",
"AI Applications"
],
"ner": {
"entities": [
{
"entity": "gte-tiny",
"type": "model"
},
{
"entity": "bge-micro",
"type": "model"
},
{
"entity": "bge-small",
"type": "model"
},
{
"entity": "all-MiniLM-L6-v2",
"type": "model"
},
{
"entity": "non-embedding params",
"type": "parameter"
}
]
}
},
"original_structure": "had_media_only"
}{
"user": {
"created_at": "2020-08-29T23:49:56.000Z",
"default_profile_image": false,
"description": "🤖 Computer scientist, next-word-prediction enjoyer\n📊 Prev. research fellow @ Stanford RegLab\n🛠️ bUiLdiNg sOmeThiNg nEw (galactic ai)\n🏳️🌈",
"fast_followers_count": 0,
"favourites_count": 5602,
"followers_count": 1639,
"friends_count": 1595,
"has_custom_timelines": true,
"is_translator": false,
"listed_count": 35,
"location": "San Francisco, CA",
"media_count": 159,
"name": "Ben (48/100)",
"normal_followers_count": 1639,
"possibly_sensitive": false,
"profile_banner_url": "https://pbs.twimg.com/profile_banners/1299856802268377090/1671821579",
"profile_image_url_https": "https://pbs.twimg.com/profile_images/1299864018081865729/CMlOyn1u_normal.jpg",
"screen_name": "andersonbcdefg",
"statuses_count": 2782,
"translator_type": "none",
"url": "https://t.co/UFbYZQueec",
"verified": false,
"withheld_in_countries": [],
"id_str": "1299856802268377090"
},
"id": "1710708732534108413",
"conversation_id": "1710708732534108413",
"full_text": "gte-tiny not small enough for you? Consider bge-micro, an embeddings model with 1/4 the layers of bge-small. It's not SOTA—compressing a model this much does incur degradation—but it's still good enough to be competitive with all-MiniLM-L6-v2, with 1/2 the non-embedding params!",
"reply_count": 7,
"retweet_count": 16,
"favorite_count": 127,
"hashtags": [],
"symbols": [],
"user_mentions": [],
"urls": [],
"media": [
{
"media_url": "https://pbs.twimg.com/media/F72n0vpb0AAA3Lf.png",
"type": "photo"
}
],
"url": "https://twitter.com/andersonbcdefg/status/1710708732534108413",
"created_at": "2023-10-07T17:28:39.000Z",
"#sort_index": "1710708732534108413",
"view_count": 66162,
"quote_count": 3,
"is_quote_tweet": false,
"is_retweet": false,
"is_pinned": false,
"is_truncated": false,
"startUrl": "https://twitter.com/andersonbcdefg/status/1710708732534108413"
}