reinforcing that the right believes a woman's only worth is around birthing/raising kids. note that Vance last night kept reinforcing *women* staying home, *women* making choices, *women* wanting childcare. because it never would occur to him that a man has an equal share of this work.
This is scary as hell. When you have mainstream media presenting these racist MAGA dreams as policy, equal to the policy ideas from Harris. This makes it so much easier for the racists to win.
This seems to be a real advance in RLHF for multi-task problems, but it's also dizzyingly complex and recursive (LLMs-as-judges are part of the formula). I miss, you know, word2vec. #machinelearningarxiv.org/pdf/2409.20370
![The Perfect Blend: Redefining RLHF with Mixture of Judges Tengyu Xu1,†, Eryk Helenowski1,†, Karthik Abinav Sankararaman1,†, Di Jin1,†, Kaiyan Peng1, Eric Han1, Shaoliang Nie1, Chen Zhu1, Hejia Zhang1, Wenxuan Zhou1, Zhouhao Zeng1, Yun He1, Karishma Mandyam1, Arya Talabzadeh1, Madian Khabsa1, Gabriel Cohen1, Yuandong Tian2, Hao Ma1, Sinong Wang1, Han Fang1 1Meta GenAI, 2FAIR, †Equal contributions arXiv:2409.20370v1 [cs.LG] 30 Sep 2024 Reinforcement learning from human feedback (RLHF) has become the leading approach for fine-tuning large language models (LLM). However, RLHF has limitations in multi-task learning (MTL) due to challenges of reward hacking and extreme multi-objective optimization (i.e., trade-off of multiple and/or sometimes conflicting objectives). Applying RLHF for MTL currently requires careful tuning of the weights for reward model and data combinations. This is often done via human intuition and does not generalize. In this work, we introduce a novel post-training pa](https://cdn.bsky.app/img/feed_fullsize/plain/did:plc:565ebob5f6hw33hjdkxty6qj/bafkreidpztkexodf3b74oapvgbyd7kdlqdqz5h4g7av24ab4wgizkhz3ji@jpeg)
Also Mormons for those who don't like to drink and prefer the highest church hierarchy to be entirely old white men. This is a church that decided the same year, but after Star Wars came out, that black people could hold equal levels of priesthood. Because "Revelation" (IRS threated tax status).
Women's wage gains mark a change in hiring dynamics, as women approach an equal share of the earnings power of all new hires. The latest from the New Hires Quality Index. 📉📈 #NHQI
There's still time to register for our virtual Disability Mentoring Day in honor of #NDEAMzoomgov.com/meeting/regi...
There's still time to register for our virtual Disability Mentoring Day in honor of #NDEAMzoomgov.com/meeting/regi...
As I grow older I find more and more things that are beautiful. There's no finite supply of it. I don't need equal supply of ugly. Beauty can grow and ugly can shrink to nothing. I realize, unfortunately, that I haven't been able to extend that grace to myself. I'm of this beautiful world, aren't I?
We just bought a pack of house number stickers for the mailbox, and I noticed that it doesn’t include equal numbers of digits, but rather a Benford-ish distribution.
Equalize, equalize, equalize ALL IS EQUAL. All SUFFERING YOU INFLICT WILL REFLECT BACK ALL YOUR RELATIONS INVITE YOU TO A WORLD OF SUFFERING Promise to lie to me and figure out it will all be in vain, you can't stop the truth of absolute reflection. Equalize before you run out of time.