Han Guo

PhD Student @MIT_CSAIL | Past: @togethercompute @LTIatCMU @MITIBMLab @UNCNLP, @SFResearch, @BaiduResearch | Machine Learning, NLP.

[Han Guo @HanGuo97]

Nice work, congrats!

[Keya Hu @HuLillian39250]

Excited to share our CVPR 2026 paper, ARC Is a Vision Problem! 🖼️

The Abstraction and Reasoning Corpus (ARC) is often approached as a language reasoning problem, despite being an inherently visual puzzle for humans.

🧩Introducing Vision ARC (VARC)🧩: we reframe abstract reasoning as an image-to-image translation problem, solved by a plain Vision Transformer.

[Keya Hu @HuLillian39250]

Excited to share our CVPR 2026 paper, ARC Is a Vision Problem! 🖼️

The Abstraction and Reasoning Corpus (ARC) is often approached as a language reasoning problem, despite being an inherently visual puzzle for humans.

🧩Introducing Vision ARC (VARC)🧩: we reframe abstract reasoning as an image-to-image translation problem, solved by a plain Vision Transformer.

[Chenfeng_X @Chenfeng_X]

Thunderagent for Dynamo!

[Simran Arora @simran_s_arora]

exciting to see our work integrated into @nvidia dynamo!

[Han Guo @HanGuo97]

@StephenLCasper @MITCSAIL @Harvard @Kennedy_School 🎉🎉congrats!!

[Edward Z. Yang @ezyang]

A common informal usage of torch.compile is to generate Triton code which people then copy paste into their codebase. https://t.co/swpryMSoL6 is an experiment to put a nice API around this workflow. Curious to see if it will get any traction!

[rishi @RishiBommasani]

AI is changing how employers hire workers.

Today we are publishing our research over the past four years into this high-stakes application of AI.

We independently studied the impacts of deployed AI hiring tools based on the real outcomes for 3.3 million people.

[Chenfeng_X @Chenfeng_X]

Maybe it is time for us to look beyond Transformers.

Some of the most important operators in future AI systems may not be attention layers. They may be classical ML primitives: KMeans, KNN, PCA, SVD, clustering, running inside fast agentic loops.

Old algorithms. New workloads. Modern hardware. Check out talented @Andy_ShuoYang 's post and blog: https://t.co/LZKv3rfWyJ

[alex zhang @a1zhang]

Introducing a minimal training harness built on prime-rl and verifiers, so you can now train your own RLMs without sandboxes! All available in the `training/` folder in the RLM GitHub repo!

We train RLM-Qwen3-30B-A3B-v0.1, using RL on a separate split of environments (OOLONG-Spam, BC+ split) to greatly improve performance across the board on long-context tasks evaluated in the original RLM paper.

We trained for a day on an 8xA100 using prime-rl; code and model are open-source and available on GitHub / Huggingface.

[Shuo Yang @Andy_ShuoYang]

Flash-KMeans was only the beginning.

Today, from the Flash-KMeans team, we are releasing FlashLib — a GPU library for fast, predictable, agent-ready classical ML operators.

Up to 26× on KMeans, 19× on KNN, 40× on HDBSCAN, 208× on TruncatedSVD, 47× on PCA, 147× on exact t-SNE, and 49× on MultinomialNB over state-of-the-art (cuML).

Blog: https://t.co/P31SGl0cyT
Code: https://t.co/9nkO2hmeOl

[Shannon Shen @shannonzshen]

"Evolve your repo, not just your agent."

Self-evolving repository (SEPO) is a fun idea that I’ve been exploring recently. @sepoagent turns any GitHub repo into a shared workspace for humans and coding agents.

[Rulin Shao @RulinShao]

DR Tulu is now accepted for an oral presentation at #ICML2026 🙏

Updated paper: https://t.co/uqUFBYqfL5
📥We added more ablations including using Qwen3-8B as the rubric generator&judge, showing evolving rubrics work with a weak model too; spurious rewards sanity check, etc.

Live demo: https://t.co/1zUsw2wBvu
Code&models: https://t.co/9hIaJW6ygY

[Ziming Mao @ziming_mao]

🚀 Excited to release mKernel: a set of fast multi-node, multi-GPU fused kernels.

💻 Code: https://t.co/y2WfdMVTfC
📝 Blog: https://t.co/wGomxmeRxr

mKernel fuses compute + communication into one persistent GPU kernel, covering both intra/inter-node with GPU-initiated communication.

Amazing team: @yangzhouy, Chon Lam Lao, Costin Raiciu, Scott Shenker, @istoica05

[Yoav Gelberg @yoav_gelberg]

Excited about this new work

As KV compaction becomes increasingly important, we ask whether it’s worth adapting the model itself to perform better under compaction

Turns out, it can really matter

[Sasha Rush @srush_nlp]

Talk: Training Composer
https://t.co/diSjKhP4g5

Overview of the methods that we use at Cursor to build our model.

[Han Guo @HanGuo97]

@Rohit_Writes Ah got it!

[Han Guo @HanGuo97]

LLM training is built on fast MatMuls. But many surrounding ops still run as memory-bound kernels.

CODA reparameterizes them to hide in the matmul’s shadow, fused into its epilogue before results leave the chip.

Bonus: LLMs can write fast CODA kernels too (approaching SoLs).

[Han Guo @HanGuo97]

@Rohit_Writes Good question, any good places to try it out?

[Shannon Shen @shannonzshen]

https://t.co/txEyWwGQ66

[Han Guo @HanGuo97]

@Tim_Dettmers Thank you Tim!

[Mingkai Deng @mdeng34]

This is a prototype using language-based world models. Stay tuned for our next steps on multimodal and physical world models.

The concept of a configurator, which decides when and how deeply to engage a reasoning process, is not specific to planning, but extensible to learning and adaptation going forward.

📄 SR²AM: https://t.co/LKeXZFN8Hh
📄 SiRA: https://t.co/5JzLSEu4nO
🌐 Project: https://t.co/1CUlEdFMxY
💻 Code: https://t.co/JSBoERYHaB

🤗 SR²AM-v0.1-8B: https://t.co/b1kkuvFL6k
🤗 SR²AM-v1.0-30B: https://t.co/PES00q6a4J

Joint work with @jinyuhou0, @larasnevess, @varad0309, @tw_killian, @waterluffy, @ericxing