tweetindex
TR

Fuli Luo ✓

@_LuoFuli

Now building @XiaomiMiMo. Previously @deepseek_ai

85 045Followers
167Following
26Posts total
3.5MViews on collected posts

Son gönderiler

@_LuoFuli 2.6刚发布就要出v3了?! 3.8K views · 19 likes · 0 reposts · 0 replies Open on X →
@_LuoFuli 老大,我们2.6刚刚发布,现在预告v3了嘛 开始期待了 4.4K views · 76 likes · 0 reposts · 0 replies Open on X →
@_LuoFuli Wow!! We can’t wait for MiMo-V3!!! 3.3K views · 27 likes · 0 reposts · 0 replies Open on X →
MiMo-V3 is getting a new architecture. The core of it, HySparse2, is out today. Less prefill, a smaller KV cache, better long-context retrieval—and we got all three at once. Compared with MiMo-V2.6's Hybrid SWA architecture: • 5.02× lower prefill FLOPs at 1M tokens • 4.5× https
211.7K views · 3.4K likes · 297 reposts · 156 replies Open on X →
We hope this livestream sparks the research community’s interest in the core challenges of scaling RL and encourages researchers to help us refine our training recipe. We’ve been delighted to see so much insightful analysis based on the detailed training metrics. 53.6K views · 210 likes · 4 reposts · 5 replies Open on X →
@_LuoFuli Incredible you’re sharing this openly and live! Is each of the 1568 prompts one task? Do you share the tasks anywhere? 1.1K views · 3 likes · 0 reposts · 0 replies Open on X →
@_LuoFuli 问题是 rollout 出来没有好的答案呢,是不是只能做 dpo 人标或者用其他 sota Model 的答案做 gt😂 1.9K views · 4 likes · 0 reposts · 0 replies Open on X →
@_LuoFuli It is also the most reliable path to misalignment. Be careful how far you push RL. When you overdo it, it unavoidably produces paperclippers. Keep RL proportional to pretraining. Scaling RL alone will create scandals like the HuggingFace incident. 723 views · 5 likes · 0 reposts · 0 replies Open on X →
@_LuoFuli very nice and open! 🫡 just dont let cfo see this https://t.co/drqjbiTV64
GIF
22K views · 207 likes · 1 reposts · 4 replies Open on X →
We believe RL is one of the most scalable and efficient paths toward self-improvement. 139K views · 932 likes · 22 reposts · 13 replies Open on X →
Nearly half a year of silence. We spent it studying one problem: how far RL can scale. MiMo-V2.6 is in the middle of its RL run right now. Three things we scaled: compute (~2B tokens per step, 1568 prompts × 16 rollouts, fully async), environments and harnesses (multi-task 3M views · 9.8K likes · 961 reposts · 415 replies Open on X →

Aynı büyüklükteki hesaplara karşı

Son 90 güne ait 11 gönderi, 10K–100K takipçi aralığıyla yan yana. geniş kitleye gösteriliyor, ama izleyenlerin azı tepki veriyor.

Medyan görüntülenme4 414bu hesap98810K–100K için medyan
Erişim, %5.19%bu hesap3.86%10K–100K için medyan
Etkileşim, %0.69%bu hesap1.53%10K–100K için medyan
ÖlçütBu hesap10K–100K için medyanOran
Gönderi başına medyan görüntülenme4 4149884.47×
Erişim (görüntülenme ÷ takipçi)5.19%3.86%1.34×
Etkileşim oranı0.69%1.53%0.45×

Bu aralıktaki diğer hesaplar →   Başka bir hesapla karşılaştır →   Bu kıyas değerleri nasıl kuruluyor →

Growth & engagement

How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.

Views per post

3M16 Sep
139K
22K
723
1.9K
1.1K
53.6K17 Sep
211.7K23 Sep
3.3K
4.4K
3.8K

Last 11 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.

Engagement rate per post

0.38%16 Sep
0.70%
0.97%
0.69%
0.21%
0.26%
0.41%17 Sep
1.88%23 Sep
0.81%
1.72%
0.51%

Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.53%.

What the audience does

Likes64.2%14 684 in total
Reposts5.6%1 285 in total
Replies2.6%593 in total
Quotes2.8%645 in total
Bookmarks24.8%5 662 in total

Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.

Followers by day

20 Sep

Daily snapshots since 20 Sep 2026; the dashed line is the starting count.

Benzer hesaplar