Preference-based LLM router training

RouteLLM learns router from preference data; routes simple queries to GPT-3.5, complex to GPT-4; maintains 95% of GPT-4 performance at 85% cost savings

?Could not be graded.0.00

Key facts

Vertical: Software & tech
Function: Other
Status: Seen in the wild
Volume: one-off
Value: none
Risk: low
Evidence: —
Flags: None

Source: https://github.com/lm-sys/RouteLLM

Build this with a classifier

Define a typed decision with a bounded answer, then evaluate it on examples.

{
  "decision_type": "choice",
  "question": "Does this input match the decision in “Preference-based LLM router training”?",
  "input": "<input to classify>",
  "output": "one label from a fixed list"
}

Related use cases

Cite this

Copy a link in your preferred format.

HTML

<a href="https://jevaiusecases.com/use-cases/preference-based-llm-router-training/">Preference-based LLM router training</a>

Markdown [Preference-based LLM router training](https://jevaiusecases.com/use-cases/preference-based-llm-router-training/)