Will RL work for LLMs "spill over" to the rest of RL by 2026?
5
1kṀ4092026
40%
chance
1D
1W
1M
ALL
RL is important for training LLMs and it seems likely that there will be significantly more investment in RL by the major LLM groups this year. Will any of the advances they make be:
Published (any publication that allows the research to be used elsewhere counts, this does not have to be a paper)
A significant advance for the rest of RL
For example, a new version of PPO that is close to SOTA for agents in Atari environments would resolve this YES.
What counts as a "significant advance" is mostly subject to my inscrutable whims, but is aimed more at cool research than important result. Think "very exciting to see at a conference" rather than "revolutionizes the field".
This question is managed and resolved by Manifold.
Get
1,000 to start trading!
1,000Related questions
Related questions
What will Manifolders mostly use LLMs for, by EOY 2025?
Will LLMs mostly overcome the Reversal Curse by the end of 2025?
60% chance
Will LLMs be able to formally verify non-trivial programs by the end of 2025?
31% chance
Will LLMs become a ubiquitous part of everyday life by June 2026?
89% chance
Will LLMs (or similar AI systems) be meaningfully integrated into US public school education by 2025?
5% chance
Will LLMs be better than typical white-collar workers on all computer tasks before 2026?
25% chance
Will an LLM improve its own ability along some important metric well beyond the best trained LLMs before 2026?
58% chance
Will LLM hallucinations be a fixed problem by the end of 2025?
14% chance
Will Apple release its own LLM on par with state of the art LLMs before 2026?
49% chance
Will one of the major LLMs be capable of continual lifelong learning (learning from inference runs) by EOY 2025?
34% chance