Scott Williams 🐧 on Nostr: "Importantly, we demonstrate that LLMs struggle even when provided with multiple ...
"Importantly, we demonstrate that LLMs struggle even when provided with multiple examples of the same question or examples containing similar irrelevant information. This suggests deeper issues in their reasoning processes that cannot be easily mitigated through few-shot learning or fine-tuning."
Published at
2024-11-07 19:38:57Event JSON
{
"id": "1edce1874b81fd1da1e905d1364ad833a74ecd8ba3596cbbfcdf70909470a084",
"pubkey": "beee28c6450eb5d21bf94279d1f3f41e524c87abf86393835b45e6a1e98390e7",
"created_at": 1731008337,
"kind": 1,
"tags": [
[
"e",
"2d9db3bc3f5d62a4037f13c0f2ca8fb161ff72330bbe77b5da03fa373691a8c4",
"wss://relay.mostr.pub",
"reply"
],
[
"proxy",
"https://mastodon.online/users/vwbusguy/statuses/113443362385850932",
"activitypub"
]
],
"content": "\"Importantly, we demonstrate that LLMs struggle even when provided with multiple examples of the same question or examples containing similar irrelevant information. This suggests deeper issues in their reasoning processes that cannot be easily mitigated through few-shot learning or fine-tuning.\"",
"sig": "e9b3a66f12c91fc576e42988bd6e2da73e1274a467619d1a923fc9aa01381577f0b982bdfa4e6094ce5b5f705155fb0b0e1d25a66082f33e940d027ffecd3d4a"
}
