scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-10-05 02:41:56 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 48

docs(router): clarify trained profile tradeoffs

Browse source
This commit is contained in:
Tin Chi Lo 2026-09-14 19:42:26 -07:00
parent 2c6c8cdfa1
commit f5870e832e
1 changed files with 1 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
cookbook/auto_router_roi_training/FINDINGS.md
View file

@ -2,7 +2,7 @@
Both implementations now support runnable, opt-in trained profiles. Training covered three cards, raw probabilities, per-model calibration, task-dependent calibration, and pair-specific boundaries, including combinations. All fitting and policy selection used the DeepSWE training and validation splits. The 25 fresh SWE-bench tasks were used only for evaluation
The selected profiles did not improve the overall quality and cost tradeoff. They remain experimental configurations, with the original defaults preserved
The selected profiles did not establish a consistent improvement over the original routers. Most paid more for the same solve count or traded away solves for savings. They remain experimental configurations, with the original defaults preserved
## Comparison with the existing routers