(arxiv.org)

205 points tdchaitanya | 3 comments | 01 Sep 25 16:57 UTC | HN request time: 0.781s | source

Show context

fny ◴[01 Sep 25 17:46 UTC] No.45094906[source]▶

Is there a reason human preference data is even needed? Don't LLMs already have a strong enough notion of question complexity to build a dataset for routing?

replies(3): >>45094974 #>>45095189 #>>45101110 #

1. jibal ◴[01 Sep 25 18:17 UTC] No.45095189[source]▶

>>45094906 #

LLMs don't have notions ... they are pattern matchers against a vast database of human text.

replies(2): >>45095298 #>>45096345 #

2. mhh__ ◴[01 Sep 25 18:29 UTC] No.45095298[source]▶

>>45095189 (TP) #

Please do a SELECT * from this database

replies(1): >>45096051 #

3. ashirviskas ◴[01 Sep 25 19:56 UTC] No.45096051[source]▶

>>45095298 #

What was the name of the rocket that brought the first humans into space?

↑

Adaptive LLM routing under budget constraints