Re:Auto-Differentiating Any LLM Workflow: A Farewell to Manu...
by xg15 on Saturday February 01, 2025 @05:14AM
(#42897330)
Just read the abstract so far. Sounds amazing, but just for the sake of understanding, what would be the inputs and outputs of such a system? If the prompt is generated, how do you tell the system what you'd like to have? And what is the ground truth that is trained against? Examples of the desired text?
[ Parent
| on HN ]
Re:Auto-Differentiating Any LLM Workflow: A Farewell to Manu...
by amelius on Saturday February 01, 2025 @08:02AM
(#42897987)
I think what they mean is intermediate prompts, i.e. the prompts that the system gives to itself when solving a problem that requires multiple stages.
Re:Auto-Differentiating Any LLM Workflow: A Farewell to Manu...
by xg15 on Saturday February 01, 2025 @09:36AM
(#42898608)
Ah, that makes sense. So (very) basically, they're putting a number of regular LLMs into a sort of compute chain/graph, where one LLM feeds into the other, then doing gradient descent on the whole chain at once, essentialy treating the boundaries between LLM n and LLM n+1 as "hidden layers"?