Log in
Log into community
Find answers from the community
View all posts
Related posts
Did this answer your question?
π
π
π
Powered by
Hall
Inactive
Updated 12 months ago
0
Follow
Also has someone else experienced that
Also has someone else experienced that
Inactive
0
Follow
k
kevingoed
12 months ago
Β·
Also has someone else experienced that
gpt-3.5-turbo-1106
produces significantly worse results as the old
gpt-3.5-turbo-16k-0613
b
k
L
7 comments
Share
Open in Discord
b
bmax
12 months ago
have seen openai forums about this
k
kevingoed
12 months ago
For my example above I get a reply that is 5 times as long
k
kevingoed
12 months ago
I also read in the OpenAI forums that people saw improvements when using new prompts, does that maybe mean Llama Index needs to update some of the hardcoded prompts?
b
bmax
12 months ago
https://community.openai.com/t/gpt-3-5-turbo-1106-is-worse-than-0613-version/519820
k
kevingoed
12 months ago
oh boy...
k
kevingoed
12 months ago
curious what @Logan M thinks about this
L
Logan M
12 months ago
hmm tbh I haven't tried the new model very much, but that doesn't surprise me at all π Classic OpenAI
Probably the prompts need to be changed slightly. But also, 1106 isn't the default model (yet) in llama-index, so I'm not too worried yet
Add a reply
Sign up and join the conversation on Discord
Join on Discord