🔥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

Nobody wants to compare with BM25 any more.

""Whenever someone asks me if RL works, I say it doesn't and 70% of the time I am right." alexirpan.com/2018/02/14/rl-…

This is a rather blockbuster piece of news: the Hugging Face library is dropping support for both Jax and Tensorflow. linkedin.com/posts/lysandre…

I don't wanna say 'I told you so', but I told you so.

I find it quite upsetting that people use the terms LLM, or PLM (pretrained LM) to denote every possible pre-trained Transformer model including BERT, T5, etc... This reminds me good old days when DNN was often used to denote a multilayer feedforward network (or at least this was

Dear ACL community, We are seeking emergency reviewers for the May cycle. Please indicate your availability (ASAP) if you can help review extra papers urgently (by the 24th of June AOE). Many thanks!

Gemini CLI is a free agent now!

arxiv.org/abs/2506.23908…

This is something I have been suspecting for a while, but I do not think we have a lot of supporting evidence though. Anyways, it looks like efficiency of generation plays an important role in user's perception of quality.

I also want to know the answer. Moreover what's the answer in the broad CS context not just AI?