🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

To train a local SWE-agent you need to synthetically generate SWE-bench-like problems+solutions. SWE-smith is our scalable solution, and we got amazing results: SWE-agent-LM 32B is an extremely strong local agent. The tutorial vids are great! John Yang youtube.com/watch?v=HO8i0T…

So cool!!

Gemini 2.5 Flash plays Final Fantasy in real-time (VideoGameBench)

As we optimize model reasoning over verifiable objectives, how does this affect human understanding of said reasoning to achieve superior collaborative outcomes? In our new preprint, we investigate human-centric model reasoning for knowledge transfer 🧵:

Congrats Dr. Merrill!

Great to see super quick and lightweight models with good SWE-bench numbers! Up until ~6 months ago these were the numbers that the frontier models were getting.

NeurIPS reviews are due on Wednesday. Thanks John Yang for reminding me :)

Do language models have algorithmic creativity? To find out, we built AlgoTune, a benchmark challenging agents to optimize 100+ algorithms like gzip compression, AES encryption and PCA. Frontier models struggle, finding only surface-level wins. Lots of headroom here!🧵⬇️

Excited to release AlgoTune!! It's a benchmark and coding agent for optimizing the runtime of numerical code 🚀 algotune.io 📚 algotune.io/paper.pdf 🤖 github.com/oripress/AlgoT… with Ofir Press Ori Press Patrick Kidger Bartolomeo Stellato Arman Zharmagambetov & many others 🧵

Im trying to edit the SWE-bench website's CSS and top models are really struggling. People please evaluate on SWE-bench Multimodal more so I don't have to spend time on our CSS. I'm begging you this is soooo boring and frustrating. Also what do you think about the redesign?

Just removing a lot of vertical space so you can see more entries without scrolling...