Yisong Yue ✓

@yisongyue · Pasadena, CA · joined 28 Jul 2009

AI Professor @Caltech (@YueLabCaltech)

24 467Followers

3 126Following

8 594Posts total

272.9KViews on collected posts

Derniers posts

Yisong Yue ✓@yisongyue7 Aug

Neurosymbolic AI is back. Can’t wait for recursive self-improvement to give us neuro(symbolicneuro)symbolic AI. 4.8K views · 34 likes · 3 reposts · 1 replies Open on X →

Yisong Yue ✓@yisongyue5 Aug

Knowledge flywheels are a new scaling dimension for self-improving AI. https://t.co/gQ4QBMtf7n 137K views · 943 likes · 95 reposts · 27 replies Open on X →

Yisong Yue ✓@yisongyue30 Jul

AI learns to improve AI, full-stack version. 2.8K views · 8 likes · 0 reposts · 0 replies Open on X →

Yisong Yue ✓@yisongyue29 Jul

Knowledge flywheels might be the most effective path to recursive self improvement. Can significantly outperform prompt and harness optimization at a fraction of the cost. 2.5K views · 8 likes · 1 reposts · 1 replies Open on X →

Yisong Yue ✓@yisongyue29 Jul

Amazing work by @AsariAILabs! Full-stack optimization of 1T+ parameter models with rigorous correctness checks. Agentic self-improvement at work. 1.8K views · 11 likes · 0 reposts · 0 replies Open on X →

Yisong Yue ✓@yisongyue29 Jul

The most interesting part? Seeing our agents learn and improve from each run. While optimizing DeepSeek v4 Pro, the agents learned how to avoid a distributed deadlock. When they later hit a similar situation optimizing GLM 5.2, they generalized that insight and saved 44 minutes

12.1K views · 28 likes · 2 reposts · 1 replies Open on X →

Yisong Yue ✓@yisongyue29 Jul

Our self-improving agents optimized the full @vLLM_project inference stack, with up to 16% more throughput and interactivity for @deepseek_ai v4 Pro and @Zai_org GLM 5.2 on B200s (no MTP). Every change was verified and our agents got better and faster at it with each iteration.

0:16

111.9K views · 253 likes · 32 reposts · 12 replies Open on X →

Face aux comptes de taille comparable

7 posts des 90 derniers jours, à côté de la tranche de 10K–100K abonnés. diffusé largement, mais peu de ces spectateurs réagissent.

Vues médianes4 820ce compte1 043médiane pour 10K–100K

Portée, %19.70%ce compte3.67%médiane pour 10K–100K

Engagement, %0.41%ce compte1.68%médiane pour 10K–100K

Indicateur	Ce compte	Médiane pour 10K–100K	Rapport
Vues médianes par post	4 820	1 043	4.62×
Portée (vues ÷ abonnés)	19.70%	3.67%	5.37×
Taux d'engagement	0.41%	1.68%	0.24×

Autres comptes de cette tranche → Comparer avec un autre compte → Comment ces repères sont établis →

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

111.9K29 Jul

12.1K

1.8K

2.5K

2.8K30 Jul

137K5 Aug

4.8K7 Aug

Last 7 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.

Engagement rate per post

0.27%29 Jul

0.26%

0.61%

0.41%

0.28%30 Jul

0.78%5 Aug

0.79%7 Aug

Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.68%.

What the audience does

Likes37.0%1 285 in total

Reposts3.8%133 in total

Replies1.2%42 in total

Bookmarks58.0%2 013 in total

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

The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.

Derniers posts

Face aux comptes de taille comparable

Growth & engagement

Views per post

Engagement rate per post

What the audience does

Comptes similaires