| ▲ | ulimn 6 hours ago | |
Isn't the outcome / solution for a given task non-deterministic? So can we reliably measure that? | ||
| ▲ | foota 6 hours ago | parent | next [-] | |
Yes, sort of. Generally you can measure the pass rate on a benchmark given a fixed compute budget. A sufficiently smart model can hit a high pass rate with fewer tokens/compute. Check out the cost efficiency on https://artificialanalysis.ai/ (say this posted here the other day, pretty neat charts!) | ||
| ▲ | torginus 6 hours ago | parent | prev | next [-] | |
This is the only correct take. The only metric that matters is cost per desired outcome. | ||
| ▲ | genericresponse 6 hours ago | parent | prev | next [-] | |
Statistically. Do many trials and measure how often it succeeds/fails. | ||
| ▲ | dns_snek 6 hours ago | parent | prev | next [-] | |
Repetition and statistics, if you have $1000++ you didn't need anyway. | ||
| ▲ | throwuxiytayq 6 hours ago | parent | prev [-] | |
It's much easier to measure a language model's intelligence than a human's because you can take as many samples as you want without affecting its knowledge. And we do measure human intelligence. | ||