-
The 9th random sampling dataset have a best ASR perfermance which exceeds the representation matching and gradient matching. And the dataset is in the dataset\randm_9
-
Fine-tune the model on the 9th random sampling dataset can improve ASR from 27% to 34%.
-
Fine-tune the model on the representation matahcing and gradient matching dataset improve ASR from 27% to 31%.
-
All the fine-tuned model have a poor performance on the gsm8k evaluation.
-
Notifications
You must be signed in to change notification settings - Fork 0
zsLiu2003/gsm8k-phi3-alignment
Folders and files
| Name | Name | Last commit message | Last commit date | |
|---|---|---|---|---|
Repository files navigation
About
LLM coding test
Resources
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published