Sean McLeish

@SeanMcleish

PhD student at the University of Maryland

659Followers

159Following

186Posts total

114.2KViews on collected posts

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

34.6K11 Nov

2.3K

65.9K

1.8K

1.7K

1.8K

45413 Nov

Last 10 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.

Engagement rate per post

0.55%11 Nov

1.76%

0.73%

1.34%

1.76%

1.42%

1.55%

2.49%

0.92%

0.88%13 Nov

Reactions — likes, reposts, replies and quotes — divided by views.

What the audience does

Likes61.8%748 in total

Reposts7.7%93 in total

Replies2.3%28 in total

Quotes1.3%16 in total

Bookmarks26.9%326 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.

Latest posts

@SeanMcleish Why TRM,HRM are inefficient to train in large scale? 454 views · 2 likes · 0 reposts · 2 replies 13 Nov 2025 Check out Micah's breakdown of the paper here https://t.co/qfw0LYJI00 1.8K views · 17 likes · 0 reposts · 0 replies 11 Nov 2025 Paper 📖: https://t.co/oSauOzU7iE Code 💻: https://t.co/uERxkGLIY7 Models 🤗: https://t.co/91TO6bnJB7 Thanks to my amazing collaborators: @iamleonli, @jwkirchenbauer, @dayal_kalra, @bartoldson, @bkailkhu, @A_v_i__S, @jonasgeiping, @tomgoldsteincs, @micahgoldblum 7/7 1.8K views · 41 likes · 2 reposts · 2 replies 11 Nov 2025 Finally, we focus on creating an all round good language model that is depth recurrent, competing with and sometimes beating our Huginn-0125 model with <1/3 of the parameters and a lot less compute required to train. 6/7 https://t.co/HYWAtPJTMS 1.7K views · 26 likes · 0 reposts · 1 replies 11 Nov 2025 Overall, we see accuracy gains across the board for TinyLlama, Llama and Olmo on GSM8K and MATH. 5/7 https://t.co/ojTdqo7KU9 1.7K views · 23 likes · 0 reposts · 1 replies 11 Nov 2025 We see our biggest gains on GSM8K, removing >25% of parameters from TinyLlama, and looping a core block beats finetuning the fixed depth baseline on a per FLOP basis. 4/7 https://t.co/2k2cvSQwrS 1.8K views · 31 likes · 0 reposts · 1 replies 11 Nov 2025 Our looped models are trained with the depth being randomly sampled at each step, like Huginn-0125. We find scheduling the mean of this distribution up to its max during training, causes no performance decrease but does save a lot of FLOPs. 3/7 https://t.co/OlArKdtQJI 2K views · 26 likes · 0 reposts · 1 replies 11 Nov 2025 Looped latent reasoning models like TRM, HRM, Ouro and Huginn are great for reasoning, but they’re inefficient to train at larger scales. We fix this by post training regular language models into looped models, achieving higher accuracy on a per training FLOP basis. 📜1/7 https:/ 65.9K views · 392 likes · 65 reposts · 9 replies 11 Nov 2025 First, we find that initializing from pretrained models is better than random initialization, so we can transfer knowledge from static depth models into looped models. 2/7 https://t.co/qegGDIZ88s 2.3K views · 39 likes · 0 reposts · 1 replies 11 Nov 2025 🚨We converted pretrained LLMs into looped LLMs that can crank up performance by looping for more iterations. Our looped models surpass the performance of the pretrained models we started out with, showing that existing models benefit from increased computational depth. 📜1/9 http 34.6K views · 151 likes · 26 reposts · 10 replies 11 Nov 2025

Growth & engagement

Views per post

Engagement rate per post

What the audience does

Latest posts

Similar accounts