@varunneal
Cautious Weight Decay is a surprisingly simple technique that has been repeatedly validated in Modded NanoGPT. I expect it will gain serious traction as the default variant of decoupled weight decay https://t.co/BLFSeqrf2w
Viewing enriched Twitter post
Cautious Weight Decay is a surprisingly simple technique that has been repeatedly validated in Modded NanoGPT. I expect it will gain serious traction as the default variant of decoupled weight decay https://t.co/BLFSeqrf2w
{
"media": [
{
"url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2011498746714943511/media_0.jpg?",
"media_url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2011498746714943511/media_0.jpg?",
"type": "photo",
"filename": "media_0.jpg"
}
],
"processed_at": "2026-01-18T17:33:34.576903",
"pipeline_version": "2.0"
} {
"type": "tweet",
"id": "2011498746714943511",
"url": "https://x.com/varunneal/status/2011498746714943511",
"twitterUrl": "https://twitter.com/varunneal/status/2011498746714943511",
"text": "Cautious Weight Decay is a surprisingly simple technique that has been repeatedly validated in Modded NanoGPT. I expect it will gain serious traction as the default variant of decoupled weight decay https://t.co/BLFSeqrf2w",
"source": "Twitter for iPhone",
"retweetCount": 14,
"replyCount": 5,
"likeCount": 288,
"quoteCount": 1,
"viewCount": 96717,
"createdAt": "Wed Jan 14 18:00:47 +0000 2026",
"lang": "en",
"bookmarkCount": 197,
"isReply": false,
"inReplyToId": null,
"conversationId": "2011498746714943511",
"displayTextRange": [
0,
198
],
"inReplyToUserId": null,
"inReplyToUsername": null,
"author": {
"type": "user",
"userName": "varunneal",
"url": "https://x.com/varunneal",
"twitterUrl": "https://twitter.com/varunneal",
"id": "3317234755",
"name": "varun",
"isVerified": false,
"isBlueVerified": true,
"verifiedType": null,
"profilePicture": "https://pbs.twimg.com/profile_images/1553295188801359872/Edi7fPKR_normal.jpg",
"coverPicture": "https://pbs.twimg.com/profile_banners/3317234755/1765895884",
"description": "",
"location": "",
"followers": 483,
"following": 536,
"status": "",
"canDm": false,
"canMediaTag": true,
"createdAt": "Sun Aug 16 22:18:19 +0000 2015",
"entities": {
"description": {
"urls": []
},
"url": {}
},
"fastFollowersCount": 0,
"favouritesCount": 25619,
"hasCustomTimelines": true,
"isTranslator": false,
"mediaCount": 116,
"statusesCount": 681,
"withheldInCountries": [],
"affiliatesHighlightedLabel": {},
"possiblySensitive": false,
"pinnedTweetIds": [
"1998224671813025849"
],
"profile_bio": {
"description": "a bridge",
"entities": {
"description": {},
"url": {
"urls": [
{
"display_url": "varunneal.github.io",
"expanded_url": "http://varunneal.github.io",
"indices": [
0,
23
],
"url": "https://t.co/Q81XSWCDwD"
}
]
}
}
},
"isAutomated": false,
"automatedBy": null
},
"extendedEntities": {
"media": [
{
"display_url": "pic.twitter.com/BLFSeqrf2w",
"expanded_url": "https://twitter.com/varunneal/status/2011498746714943511/photo/1",
"ext_media_availability": {
"status": "Available"
},
"features": {
"large": {},
"orig": {}
},
"id_str": "2011498590095753217",
"indices": [
199,
222
],
"media_key": "3_2011498590095753217",
"media_results": {
"id": "QXBpTWVkaWFSZXN1bHRzOgwAAQoAARvqR09w20ABCgACG+pHc+gWcBcAAA==",
"result": {
"__typename": "ApiMedia",
"id": "QXBpTWVkaWE6DAABCgABG+pHT3DbQAEKAAIb6kdz6BZwFwAA",
"media_key": "3_2011498590095753217"
}
},
"media_url_https": "https://pbs.twimg.com/media/G-pHT3DbQAE71Vj.jpg",
"original_info": {
"focus_rects": [
{
"h": 238,
"w": 425,
"x": 348,
"y": 0
},
{
"h": 238,
"w": 238,
"x": 441,
"y": 0
},
{
"h": 238,
"w": 209,
"x": 456,
"y": 0
},
{
"h": 238,
"w": 119,
"x": 501,
"y": 0
},
{
"h": 238,
"w": 1500,
"x": 0,
"y": 0
}
],
"height": 238,
"width": 1500
},
"sizes": {
"large": {
"h": 238,
"w": 1500
}
},
"type": "photo",
"url": "https://t.co/BLFSeqrf2w"
}
]
},
"card": null,
"place": {},
"entities": {},
"quoted_tweet": {
"type": "tweet",
"id": "2011493120551334249",
"url": "https://x.com/lzchen_ut/status/2011493120551334249",
"twitterUrl": "https://twitter.com/lzchen_ut/status/2011493120551334249",
"text": "modded-nanogpt WRs 🤝 Cautious Weight Decay\n\nPretty much all the @speedrun WRs since November have used CWD in some form (e.g., @varunneal's \"CWD w/ schedule\"). \n\nHuge thanks to everyone experimenting and sharing results — shoutout to @varunneal, @classiclarryd, @ChrisJMcCormick, @roeeshenberg, @YouJiacheng, and the other contributors.\n\nPaper 👇 https://t.co/0qWhQnUuLv",
"source": "Twitter for iPhone",
"retweetCount": 5,
"replyCount": 2,
"likeCount": 47,
"quoteCount": 1,
"viewCount": 75102,
"createdAt": "Wed Jan 14 17:38:26 +0000 2026",
"lang": "en",
"bookmarkCount": 43,
"isReply": false,
"inReplyToId": null,
"conversationId": "2011493120551334249",
"displayTextRange": [
0,
304
],
"inReplyToUserId": null,
"inReplyToUsername": null,
"author": {
"type": "user",
"userName": "lzchen_ut",
"url": "https://x.com/lzchen_ut",
"twitterUrl": "https://twitter.com/lzchen_ut",
"id": "1194040130249510912",
"name": "Lizhang Chen",
"isVerified": false,
"isBlueVerified": true,
"verifiedType": null,
"profilePicture": "https://pbs.twimg.com/profile_images/1870198712582189057/dzUsJSBe_normal.jpg",
"coverPicture": "https://pbs.twimg.com/profile_banners/1194040130249510912/1698816449",
"description": "",
"location": "Austin, TX",
"followers": 142,
"following": 362,
"status": "",
"canDm": true,
"canMediaTag": true,
"createdAt": "Mon Nov 11 23:53:19 +0000 2019",
"entities": {
"description": {
"urls": []
},
"url": {}
},
"fastFollowersCount": 0,
"favouritesCount": 145,
"hasCustomTimelines": true,
"isTranslator": false,
"mediaCount": 4,
"statusesCount": 94,
"withheldInCountries": [],
"affiliatesHighlightedLabel": {},
"possiblySensitive": false,
"pinnedTweetIds": [
"2011493120551334249"
],
"profile_bio": {
"description": "Student researcher @Google @GoogleDeepMind | Ph.D student at UT Austin",
"entities": {
"description": {
"user_mentions": [
{
"id_str": "0",
"indices": [
19,
26
],
"name": "",
"screen_name": "Google"
},
{
"id_str": "0",
"indices": [
27,
42
],
"name": "",
"screen_name": "GoogleDeepMind"
}
]
},
"url": {
"urls": [
{
"display_url": "l-z-chen.github.io",
"expanded_url": "https://l-z-chen.github.io/",
"indices": [
0,
23
],
"url": "https://t.co/W3aiLaHIka"
}
]
}
}
},
"isAutomated": false,
"automatedBy": null
},
"extendedEntities": {
"media": [
{
"display_url": "pic.twitter.com/yctEhrSJyP",
"expanded_url": "https://twitter.com/Tim38463182/status/2011493120551334249/photo/1",
"ext_media_availability": {
"status": "Available"
},
"features": {
"large": {},
"orig": {}
},
"id_str": "2011493110375989254",
"indices": [
305,
328
],
"media_key": "3_2011493110375989254",
"media_results": {
"id": "QXBpTWVkaWFSZXN1bHRzOgwAAQoAARvqQlOYG0AGCgACG+pCVfaasWkAAA==",
"result": {
"__typename": "ApiMedia",
"id": "QXBpTWVkaWE6DAABCgABG+pCU5gbQAYKAAIb6kJV9pqxaQAA",
"media_key": "3_2011493110375989254"
}
},
"media_url_https": "https://pbs.twimg.com/media/G-pCU5gbQAYG-Y5.jpg",
"original_info": {
"focus_rects": [
{
"h": 559,
"w": 998,
"x": 0,
"y": 0
},
{
"h": 559,
"w": 559,
"x": 0,
"y": 0
},
{
"h": 559,
"w": 490,
"x": 0,
"y": 0
},
{
"h": 559,
"w": 280,
"x": 38,
"y": 0
},
{
"h": 559,
"w": 1024,
"x": 0,
"y": 0
}
],
"height": 559,
"width": 1024
},
"sizes": {
"large": {
"h": 559,
"w": 1024
}
},
"type": "photo",
"url": "https://t.co/yctEhrSJyP"
}
]
},
"card": null,
"place": {},
"entities": {
"urls": [
{
"display_url": "arxiv.org/abs/2510.12402",
"expanded_url": "https://arxiv.org/abs/2510.12402",
"indices": [
346,
369
],
"url": "https://t.co/0qWhQnUuLv"
}
],
"user_mentions": [
{
"id_str": "1760843364386443264",
"indices": [
64,
73
],
"name": "a16z speedrun 🧊",
"screen_name": "speedrun"
},
{
"id_str": "3317234755",
"indices": [
127,
137
],
"name": "varun",
"screen_name": "varunneal"
},
{
"id_str": "3317234755",
"indices": [
234,
244
],
"name": "varun",
"screen_name": "varunneal"
},
{
"id_str": "1791834924317560832",
"indices": [
246,
260
],
"name": "Larry Dial",
"screen_name": "classiclarryd"
},
{
"id_str": "33635019",
"indices": [
262,
278
],
"name": "Chris McCormick",
"screen_name": "ChrisJMcCormick"
},
{
"id_str": "500099892",
"indices": [
280,
293
],
"name": "Roee Shenberg",
"screen_name": "roeeshenberg"
},
{
"id_str": "3426190841",
"indices": [
295,
307
],
"name": "You Jiacheng",
"screen_name": "YouJiacheng"
}
]
},
"quoted_tweet": null,
"retweeted_tweet": null,
"isLimitedReply": false,
"article": null
},
"retweeted_tweet": null,
"isLimitedReply": false,
"article": null
}