🔥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

We enabled OLMoTrace for Tülu 3 models! 🤠 Matched spans are shorter than for OLMo models, bc we can only search in Tülu's post-training data (base model is Llama). Yet we thought it'd still bring some value. Try yourself on the Ai2 playground -- playground.allenai.org

Only a fraction of data needed for LLM comes with identifiable licenses. But if you curate it all, can you train a model on in? We release Common Pile, a 1T tokens dataset, and train a 7B model on it! results are on par with open weights models trained on eq FLOPS

does anyone **actually** read the output of research assistant or do you also immediately go to references

one hard lesson i’ve learned on working on large scale ML system: no change is too small to be ablated. every hypothesis consumes a non-negligible amount of bandwidth, and there are simply Too Many Things to try. instead of arguing whether something should work, just try it

Google Cloud somewhat borked, and has brought WandB down with it. Might as well pack and go home?

probably the greatest paper title in history

So glad to see Molmo recognized. Amazing work lead by Chris, Matt Deitke, and Ani Kembhavi; I feel so fortunate to have at least helped a lil bit!

What happens when an LLM is asked to use information that contradicts its knowledge? We explore knowledge conflict in a new preprint📑 TLDR: Performance drops, and this could affect the overall performance of LLMs in model-based evaluation.📑🧵⬇️ 1/8 #NLProc #LLM #AIResearch

so cool that it points to bit.ly/3ZD47Xw

this is really great work.

literally the only bad thing about uv is that some ppl are STILL not using uv 😒

invite me to your office and feed me and i will stick your logo on my laptop