In the paper Absolute Zero: Reinforced Self-play Reasoning with Zero Data
https://arxiv.org/pdf/2505.03335
AI learns from scratch on its own to create and prove problems of reasoning incl. deduction, abduction and induction for coding or mathematics in the manner of Alpha(Go)Zero or KataGo creating and improving go decision-making.
Absolute Zero
-
RobertJasiek
- Judan
- Posts: 6272
- Joined: Tue Apr 27, 2010 8:54 pm
- GD Posts: 0
- Been thanked: 797 times
- Contact:
-
xela
- Lives in gote
- Posts: 652
- Joined: Sun Feb 09, 2014 4:46 am
- Rank: Australian 3 dan
- GD Posts: 200
- Location: Adelaide, South Australia
- Has thanked: 219 times
- Been thanked: 281 times
Re: Absolute Zero
Intriguing... It's very jargony, more so than the various AlphaGo papers. I can't tell if this software is actually doing something brilliant, or if it's merely an improvement on previous models that were terrible.
By the way, Robert's link is the direct link to the PDF download. If you want to read the abstract before deciding whether to download the full paper, go to https://www.arxiv.org/abs/2505.03335
By the way, Robert's link is the direct link to the PDF download. If you want to read the abstract before deciding whether to download the full paper, go to https://www.arxiv.org/abs/2505.03335