Elias @notes_own
The past is closer to the future than now 支那猪滚 天堂下水道 Joined June 2011-
Tweets22K
-
Followers240
-
Following3K
-
Likes19K
🚨 NuRL: Nudging the Boundaries of LLM Reasoning GRPO improves LLM reasoning, but often within the model's "comfort zone": hard samples (w/ 0% pass rate) remain unsolvable and contribute zero learning signals. In NuRL, we show that "nudging" the LLM with self-generated hints…
Tinker is cool. If you're a researcher/developer, tinker dramatically simplifies LLM post-training. You retain 90% of algorithmic creative control (usually related to data, loss function, the algorithm) while tinker handles the hard parts that you usually want to touch much less…
Tinker is cool. If you're a researcher/developer, tinker dramatically simplifies LLM post-training. You retain 90% of algorithmic creative control (usually related to data, loss function, the algorithm) while tinker handles the hard parts that you usually want to touch much less…
A missing link between Transformers and the brain? 🧠 Dragon Hatchling (BDH) is a new LLM architecture based on a scale-free, biologically-inspired network of locally-interacting neuron particles. It rivals GPT2 performance, but is designed for interpretability.
Success of RL post-training hinges on the quality of generated rollouts, but high-reward targets are sparsely scattered in the vast state space, hindering the effectiveness of reward optimization💫. 🧩Solution? 💡𝐒𝐞𝐚𝐫𝐜𝐡-𝐭𝐲𝐩𝐞 Inference-time Scaling +…
perhaps this is what so-called frontier labs do: RL before KD. 🧐 stay tuned for a detailed 🧵 from @_sungmin_cha ! the preprint link ⬇️
Exploration is fundamental to RL. Yet policy gradient methods often collapse: during training they fail to explore broadly, and converge into narrow, easily exploitable behaviors. The result is poor generalization, limited gains from test-time scaling, and brittleness on tasks…
TruthRL Incentivizing Truthful LLMs via Reinforcement Learning
🚀Ever wondered how to make RL work on impossible hard tasks where pass@k = 0%? 🤔 In our new work, we share the RL Grokking Recipe: a training recipe that enables LLMs to solve previously unsolvable coding problems! I will be at #CoLM2025 next week so happy to chat about it!…
Even with full-batch gradients, DL optimizers defy classical optimization theory, as they operate at the *edge of stability.* With @alex_damian_, we introduce "central flows": a theoretical tool to analyze these dynamics that makes accurate quantitative predictions on real NNs.
what a beautiful theory!
what a beautiful theory! https://t.co/5dGHnNJDmL
Are you ready for web-scale pre-training with RL ? 🚀 🔥 New paper: RLP : Reinforcement Learning Pre‑training We flip the usual recipe for reasoning LLMs: instead of saving RL for post‑training, we bring exploration into pretraining. Core idea: treat chain‑of‑thought as an…
The most effective way to achieve better performance is through pre-training of RL. This unlocks a lot of high-quality data. Right now, pretraining on graduate physics or maths texts is allowed the same compute as text with low information density. The model cannot predict…
The most effective way to achieve better performance is through pre-training of RL. This unlocks a lot of high-quality data. Right now, pretraining on graduate physics or maths texts is allowed the same compute as text with low information density. The model cannot predict…
The biggest news of the day: John Schulman has dropped a new blog post.
The biggest news of the day: John Schulman has dropped a new blog post. https://t.co/zriy4Kiz1J
Sonnet 4.5 is out! It’s the most aligned frontier model yet; a lot of progress relative to Sonnet 4 and Opus 4.1!
Pretraining Large Language Models with NVFP4 "We validate our approach by training a 12-billion-parameter model on 10 trillion tokens -- the longest publicly documented training run in 4-bit precision to date. Our results show that the model trained with our NVFP4-based…
What happens when your verifier decides what your model can (and can't) learn? We've been digging into this for a while, and we're excited to finally share our findings 🧵
RLHI: Reinforcement Learning from Human Interaction • Moves beyond expert-annotated data → learns from real user conversations • Two methods: 1. User-Guided Rewrites 2. User-Based Rewards • Outperforms baselines in personalization, instruction-following & reasoning
🌀New work: Era of Real-World Human Interaction 🌀 📝: arxiv.org/abs/2509.25137 - RL *directly* from User Conversations - Organic replies + long-term history are learning signal - Trained on WildChat, beats RLHF at *user* level -> the future for personal Super Intelligence? 🧵1/6
ReasoningBank: memory for self-evolving LLM agents • Distills strategies from both successes & failures • Enables agents to learn, reuse, and improve over time • Outperforms prior memory methods on web & SWE tasks (+34.2% eff., –16% steps)
Step-by-Step Diffusion: An Elementary Tutorial
Mooxau @Mooxau057
10 Followers 2K Following
Juliana @Julianareagann
150 Followers 3K Following I PLEDGE ALLEGIANCE TO THE FLAG OF THE UNITED STATES OF AMERICA 🇺🇸
RobertaNorth @ltus13PNMc99C
158 Followers 2K Following
Pitt#0321 @Miezul
116 Followers 751 Following Hard work is not just a hot-headed obsession. Knowing when to give up may be more valuable.
LetitiaHerty @rGL4S71eNh82t
19 Followers 598 Following
Fuck Commie✝️🇺... @Price9282416611
596 Followers 2K Following 反共反女權反LGBTQ反DEI/Christian/亞細亞主義/超國家主義/錫安主義/三民主義/新自由主義/第三條道路/中間至中間偏左（部分觀點為右派）/俄羅斯是歐洲最後的良心和希望
rust_die @rust_die
18 Followers 1K Following
--- @hd2432778574039
3 Followers 127 Following
绿山工作室出手... @GNghiem66286
7 Followers 100 Following 主营：国内外手机卡 流量卡注册卡 出售；微信 抖音 支付宝 钉钉 探探 陌陌 快手 成品号 等等业务 #国内注册卡!#流量卡# 香港通话卡 #全功能卡 #国外纯注册卡，所有卡均可在国内正常直接使用! 无需实名! 电报✈️：@Lai66691 【绿山工作室】
SmythOS @Smyth_OS
303 Followers 4K Following The First Open Source AI Operating System Powered by a Network of Coordinated AI Agents. Welcome to the Future of Enterprise AI.
Fiona Sit 薛凱琪 @physit33
142 Followers 7K Following WE EARN & WE LIVE TO GIVE ❤️😌Singer/Actress/Director/Fashion brand founder
アイオワ @BB61IowaUss
1K Followers 5K Following Hi！MeがIowa級戦艦、Iowaよ。Youがこの艦隊のAdmiralなの？いいじやない！私たちのこともよろしく！
Lucello葉 @fangtiago
43 Followers 893 Following INFP | Conservative rules for myself | Liberal views for the others | seeker of truth | lover of dogs | student of many others | 朋友圈外圈
xjp99ply @xjp99ply80658
44 Followers 2K Following
EV1L D3M0 @D3m0Ev1l
2K Followers 7K Following Pansexual femboy/demiboy Hong Kong INFP-T ADHD diagnosed 3D animator Native Cantonese speaker Speaks Cantonese, Mandarin, English LGBTQIA+friendly 🏳️🌈 🏳️⚧️
雪崩Avalanche @OGVR6Cg2Hw58368
75 Followers 1K Following
司马南 @jin_kong7787
124 Followers 2K Following
Peter Ye @PeterYeArizona
734 Followers 4K Following
Sophia Wu (吳雅婷) @sophaiwu1
25 Followers 413 Following ✨ 美籍华人 | 🌏 在推文中搭建文化桥梁 | ☕ 茶永远胜过咖啡 | 科技 + 旅行 + 思考 | 观点仅代表个人，但通常是对的 😉
Huajun. @rennrennrinnrin
91 Followers 4K Following
Sramir @Sramir500
66 Followers 3K Following
AI Tools Network @aitoolsnetwork
342 Followers 4K Following an online hub to find the the best AI tools
MissJoyCloud @Zrisui910
29 Followers 2K Following "Every day is a chance to make a difference, even in the smallest of ways."
SleepWellStocks🇺�... @Yxapeh7388305
57 Followers 2K Following 15-30% Monthly | 2 High-Conviction Stocks.Short-Term Gains: 15-20% in Days/Weeks.DM "JOIN" for WhatsApp Alerts. Live Trade Signals • Market Analysis
stanlea harris @stanleax
26 Followers 226 Following
阿瞒波娜娜 @amanamanbanana
12 Followers 52 Following
tang @zhunzzhhuunn
1 Followers 82 Following
Varjea @Varjea5867
28 Followers 1K Following
Gianni Dickens @gianni11758
26 Followers 2K Following
面子とよくわか... @sa2679184621917
142 Followers 560 Following 支那人はある意味では、妄想の中で生きているかわいそうで救いようのない生命体というのがよくわかる。 かかわってはいけない生き物。
mjygb263 @mjygb263
32 Followers 1K Following
専業投資家ヤマ... @CamilleWil81762
8 Followers 462 Following
献忠planB（支人�... @xiolnq21
188 Followers 2K Following 人民警察大学 情报指挥学专业毕业 /中共统战、情治及认知作战策略研究 有法律执业资格 主攻不良资产处置 / 中共代表绝大多数中国人/ 真理就是规律，规律就是上帝的律法。/ 人没资格检验上帝/人不是检验真理的标准/拜人教拜物教国家即是地狱 兜兜转转走走看看 思辨中去伪存真 互fo 公平交流就行/或激情互骂就好了
刘超 @Ericcl8964
81 Followers 3K Following
香菜 @XiangCai35015
3 Followers 165 Following
烟雨战神 @yan_shen85469
77 Followers 1K Following
hedoήist @hedo_ist
18K Followers 714 Following obsessive creator, conceptual art (original creations) Alex Iché
🐳 @Lotscheap1
2K Followers 211 Following
彩葉（いろは） @iroha250412
2K Followers 336 Following 裏垢🔞29、既婚 見られたがり性欲お化け フォロー・いいね・リポスト喜びます💞 DMしません会えません 写真の場所は文字にしないでね なりすましさん発生⚠️ご注意ください
Anime Vibes @finalformlab
252K Followers 23 Following Curated Anime Content | Turn on notifications 🔔
ObtainerOfRareAntiqui... @ObtainerOf
27K Followers 501 Following Classics-Anthropology-Antiquities. Fan of adventure, weird pieces of history, culture and folklore.
🎞️ shitposts.mp4... @shitposts_mp4
197K Followers 240 Following shitposts 💩 posted every hour, follow me if you laugh | dm for removal ✉️
Reihoe @howlongondaboat
406 Followers 207 Following
Retro Anime Screencap... @RetroScreencaps
50K Followers 16 Following Unedited screenshots from The Golden Age of Anime 🎌 No weird filters, no fake subtitles ❌ A Japanese journey through the 80s & 90s 📺 (and sometimes 70s) ✌
videogame stores @storesfromvidya
71K Followers 803 Following shops, stores and businesses from videogames. mostly loop GIFs. feel free to send a DM for submissions.
CRT Bot2 @crt_bot2
122K Followers 111 Following Just a bot that posts CRT TVs and monitors. DM for removal or submissions. Banner by @castle_zotz
面条喵🍥🐱🐈 @MTNEKO8964
2K Followers 2K Following MtX🍥 / 生产日期：8964.07.13 / hrt：25.2.11 / 55cm 169kg｜极度社恐 / 泛性恋 / 阿斯伯格 / 一只努力学习做好人类的猫｜面条的tg频道https://t.co/6TaiTWHzEF｜我的tg@MTNEKO123｜对象🥺🍜❤️@sanmiguelnoodle
𝐋𝐨𝐰 𝐏𝐨... @PolyDepression
148K Followers 405 Following | Vohyak | 📼 90s & 00s Gaming Creator | Curator | Artist 🌴 OC and retro preservation 💾 @GOGcom Partner 📩 [email protected]
Jake Rizzbot @jake_rizzbot
93 Followers 3 Following 0x7ce5314DeA0420E4d3F5aef9Cf93A277Fa0b0747 Jake the Rizzbot is smooth-talking robot wingman on STORY
Rizzbot @rizzbotofficiaI
10 Followers 0 Following Rizzing, roasting, no filter. Daily life of a real humanoid 🤖
Pretty Cities @PrettyCitiesX
76K Followers 6K Following Sharing Travel Tales | Some of the Posted content isn't mine | Please DM for Credit or Removal 🙏
🏛 𝐒𝐭𝐞𝐯... @nonregemesse
39K Followers 39 Following Strategos of the Twittercon. Degrees in Prehistoric & Roman Archaeology, & Law. Enjoyer of Roman & Medieval history.
Carlos That Notices T... @QuetzalPhoenix
47K Followers 4K Following The way of the world is to bloom and to flower and die but in the affairs of men there is no waning and the noon of his expression signals the onset of night.
Francisco Ribeiro @fraveris
185K Followers 2K Following
central dogma special... @takimunk
5K Followers 616 Following savoring the last moments before ASI | building
Philosophy Of Physics @PhilosophyOfPhy
9K Followers 0 Following Tweets about Physics, Math, Philosophy and Quotes from History. Portrait ©Sir Issac Newton
気になるニュー... @penpen_popnews
11K Followers 135 Following 2025/6/27参院選中にYoutube永久BAN→2025/8/12復活 配信とは違うｘ版ポスト投稿 モットー「他人にビジネス保守と言うなら、お前が広告を外してから言え」
在華坊 @zaikabou
27K Followers 4K Following 関心領域は酒、食べ物、美術、旅、横浜、建築、演芸などなど。横浜を中心に国内をよく移動しています。はてなブログで日記書いてます https://t.co/c7lG6VaOF3
ロアネア@最多�... @roaneatan
121K Followers 21K Following 基本、国内初の情報のみを最速で（再紹介あり）。他の人とかぶらないよう既出やどこかで見たようなのも回避。別垢でも紹介（@roaneatan2）。名前を真似てきてる大キショという業者とは無関係。DMや返信など長期的に放置中です、すみません。ほぼbot状態です、TL見れてません。
令和速報〜trendi... @reizisok
5K Followers 7K Following マスメディアが載せない報道しない、ありのままの真実の時事ネタ・ニュースに関するポストをします📮日本のことが大好きな方フォロー歓迎です！
PizzaDares & Public F... @PizzaFlashing
697K Followers 665 Following I try to find the BEST Public Nudity & Pizza Dares. Want to participate? DM! I do not own nor claim to own any videos or images posted. DM for take-downs/credit
Computer ♥ Records @ComputerLove_
39K Followers 26 Following Underground Synth Instrumental Record Label based in LA, inspired by 80s nostalgia. Check out the latest releases: https://t.co/PlJ638mmrM
The Golden Days @TheGoldenDays
20K Followers 3K Following where nostalgia lives and the world stays nostalgic from games to tech and beyond.
ヒロクライム�... @tannokasa3
60K Followers 24K Following 主に食べ物動画の紹介をしています✨ 釣りとお酒が趣味の普通の会社員です🎣 🌈他のアカウントでも色々と紹介しています🌈 @tannokasa4 ➡面白動画 @tannokasa5 ➡日常ポスト @tannokasa6 ➡動物動画 @tannokasa9 ➡動物動画
ツイッター速報... @buzsokk
187K Followers 26K Following 今話題のツイート・ニュースをネットの声と共にいち早く分かりやすくWeb記事にします。※ポストは過去の出来事も含みます。動画は規約に従い全て引用する形で投稿。フォローバックは順次通知から3日以内にしています。（DMがスパムで荒れているので3日以上返信ない場合はスタンプをお願いします）
素晴らしい世界... @yabaaata
78K Followers 1K Following World information🇯🇵 I hope one day we can see a more peaceful world.平和な日本に感謝🙏🥲
Out of context Cathol... @nocontextcathol
27K Followers 0 Following DM request | [email protected] | ID BINANCE: 123657548 | 1MCvEKU3Za9p8HhbKzwXz3DYj8CyxLNXgM
Mambo Italiano @mamboitaliano__
182K Followers 304 Following Hidden gems from the most beautiful country in the world: Italy 🇮🇹
Trad West @trad_west_
378K Followers 162 Following 🌍 Western Civilization 🌍 👑✝️ Ave Christus Rex ☦️👑
Piotr Binkowski @piotrbinkowski
44K Followers 7K Following exploring futures that never were 🚀✨ bite-sized stories, retro sci-fi and beyond + AI CPP @AIVideoDotCom up for collab✌
Samuel Smith @samuelsmith_art
18K Followers 254 Following freelance visdev artist for animated films, learning gamedev slowly ✉️[email protected]
God Save Great Britai... @GSGB01
133K Followers 6K Following Protecting British Culture and Heritage 🇬🇧 Exposing the Truth 🇬🇧
bycloud @bycloudai
9K Followers 713 Following I make youtube vids on cool AI research /// AI papers newsletter https://t.co/Xn7GMDbQSd /// paper recap @TheAITimeline /// building @findmypapersAI
Grummz @Grummz
272K Followers 4K Following Mark Kern, CEO & Designer. Former Team Lead for OG World of Warcraft. Producer, Diablo 2, Starcraft. Game lead, Firefall creator. Chrono Trigger is best game.
Hero Of Justice @herooffjustice
4K Followers 486 Following Learning AI/ML, Messiah? Soon. Hero Of Justice ⚔️
Trends for United States
You might like
