08-24-2026, 11:56 AM
Tried to reproduce it just now. The issue went away once I changed one setting and restarted the stack cleanly. The main thing that moved was latency, not quality.
measure twice, benchmark three times
Inference
›
Evaluation & Testing
›
Benchmarks
comparing models without fooling yourself: early impressions
comparing models without fooling yourself: early impressions
|
comparing models without fooling yourself: early impressions
|
|
« Next Oldest | Next Newest »
|
| Messages In This Thread |
|
comparing models without fooling yourself: early impressions - by benchplease - 08-24-2026, 11:31 AM
Re: comparing models without fooling yourself: early impressions - by cacheline - 08-24-2026, 11:56 AM
|
Users browsing this thread: 1 Guest(s)