News for hackers. Stuff that matters.

A comment on The wall confronting large language models. Parent comment

Re:The wall confronting large language models by skissane on Wednesday September 03, 2025 @08:22PM (#45121936)
LLM backtracking is an active area of research, see e.g.

https://arxiv.org/html/2502.04404v1

https://arxiv.org/abs/2306.05426

And I was wrong that nobody has implemented it, as these papers prove people have… it is just the results haven’t been sufficiently impressive to support the transition from the research lab to industrial use - or at least, not yet

[ Parent | on HN ]
Re:The wall confronting large language models by measurablefunc on Wednesday September 03, 2025 @08:31PM (#45122006)
> Empirical evaluations demonstrate that our proposal significantly enhances the reasoning capabilities of LLMs, achieving a performance gain of over 40% compared to the optimal-path supervised fine-tuning method.
[ Parent | on HN ]
Re:The wall confronting large language models by afiori on Wednesday September 03, 2025 @10:54PM (#45122999)
I would expect to see something like this soonish as around now we are seeing the end of training scaling and the beginning of inference scaling
[ Parent | on HN ]
Re:The wall confronting large language models by foota on Wednesday September 03, 2025 @11:24PM (#45123196)
This is a neat observation, training has been optimized to hell and inference is just beginning.
[ Parent | on HN ]