jb55 on Nostr: You can completely game them by training on the benchmarks using RLHF. Considering ...
You can completely game them by training on the benchmarks using RLHF. Considering now that we know elon blatantly lies about things i wouldn’t be that surprised.
Published at
2025-02-22 15:46:07Event JSON
{
"id": "90384b325e14ad9d2e94a8bcf38d878b53b93d2ed60ed770130e94409cf369f6",
"pubkey": "32e1827635450ebb3c5a7d12c1f8e7b2b514439ac10a67eef3d9fd9c5c68e245",
"created_at": 1740239167,
"kind": 1,
"tags": [
[
"e",
"6c4d9cd6bcf01f2d50b05b8a99f8d2b4c736679d89444aa5620327761c5e647a",
"",
"root"
],
[
"p",
"c6f7077f1699d50cf92a9652bfebffac05fc6842b9ee391089d959b8ad5d48fd"
]
],
"content": "You can completely game them by training on the benchmarks using RLHF. Considering now that we know elon blatantly lies about things i wouldn’t be that surprised.",
"sig": "cbb6d51806cb55205fad643a0c9416ce99154741add8b894bd39333f1a2df067f12553bc73112c8c42fe9cd7f48f5586933f62ddd33f62c05af01490117f90a4"
}
