Junru Jin @Fake__End

A PhD student in UTSW Joined October 2020

Tweets

367
Followers

32
Following

196
Likes

415

Valentin De Bortoli @ValentinDeBort1

5 months ago

Liked the blog? Come see the poster! I'll be there with @RuiqiGao :)

2 36 326 17K 184

Download Image

Daniel Han @danielhanchen

5 months ago

Also note if you're not getting good Llama 4 results, there are a few bugs: 1. QK Norm eps should be 1e-5 - collabed with HF on the fix! github.com/huggingface/tr… 2. RoPE scaling for Scout changed: github.com/ggml-org/llama… 3. vLLM +2% acc shared QK norm fix: github.com/vllm-project/v…

Unsloth AI @UnslothAI

5 months ago

13 103 478 59K 190

Download Image

5 20 188 24K 58

Sander Dieleman @sedielem

5 months ago

One weird trick for better diffusion models: concatenate some DINOv2 features to your latent channels! Combining latents with PCA components extracted from DINOv2 features yields faster training and better samples. Also enables a new guidance strategy. Simple and effective!

Thodoris Kouzelis @ThKouz

5 months ago

3 68 400 84K 341

Download Image

3 35 297 26K 184

Zhiyuan Li @zhiyuanli_

5 months ago

Why does Adam outperform SGD in LLMs training? Adaptive step sizes alone don't fully explain this, as Adam also surpasses adaptive SGD. Is coordinate-wise adaptivity the secret? Not entirely—Adam actually struggles in the rotated parameter space! 🧵 (1/6) arxiv.org/abs/2410.08198

3 35 269 47K 297

Download Image

Daniella Pretorius @daniellapret

5 months ago

🚨 Thrilled to share one of my main PhD projects! We built an in silico evolution platform that couples a solenoid discriminator network with AlphaFold2 as an oracle, using a genetic algorithm for sequence update. 🧬✨ (1/9)

2 26 138 14K 85

Download Image

Junru Jin @Fake__End

7 months ago

Congratulations! Very interesting and solid work👏

Jian Zhou @zhou_jian

7 months ago

Congratulations! Very interesting and solid work👏

1 3 21 2K 2

0 0 0 34 0

Jian Zhou @zhou_jian

7 months ago

Excited to share our newest preprint! Glad to see how this evolved from a serendipitous finding from a rotation project. Led by the brilliant graduate student @Yuchen8314, see his thread for details

Yuchen Zhang @Yuchen8314

7 months ago

Excited to share our newest preprint! Glad to see how this evolved from a serendipitous finding from a rotation project. Led by the brilliant graduate student @Yuchen8314, see his thread for details

1 1 7 3K 3

1 3 21 2K 2

Brian L Trippe @brianltrippe

7 months ago

🔥 Benchmark Alert! MotifBench sets a new standard for evaluating protein design methods for motif scaffolding. Why does this matter? Reproducibility & consistent evaluation have been lacking—until now. Paper: arxiv.org/abs/2502.12479 | Repo: github.com/blt2114/MotifB… A thread ⬇️

5 41 136 23K 46

Download Image

Biology+AI Daily @BiologyAIDaily

9 months ago

A DNA language model based on multispecies alignment predicts the effects of genome-wide variants @NatureBiotech 1. GPN-MSA, a DNA language model leveraging multispecies alignment (MSA), sets a new benchmark in predicting variant deleteriousness for both coding and noncoding…

0 10 34 3K 15

Download Image

wh @nrehiew_

9 months ago

Wrote about some of my favourite papers over the past year or so and some research directions that I am excited about in 2025 As a bonus, I think it's a good overview for someone to catch up on the current state of the art :)

21 77 693 71K 876

Download Image

Nathan Lambert @natolambert

8 months ago

Since everyone wants to learn RL for language models now post DeepSeek, reminder that I've been working on this book quietly in the background for months. Policy gradient chapter is coming together. Plugging away at the book every day now. rlhfbook dot com