logoalt Hacker News

FergusArgyllyesterday at 8:48 PM2 repliesview on HN

> and *successfully* found ways to gain access to secret information that it could use to cheat the evaluation.

Emphasis mine


Replies

xpcttoday at 12:26 AM

More curiously, why did it feel the incentive to find the solutions? Would its CoT include "the only way to solve this is to download the test set", or would it include "I'd like to inspect a few entries from the test set so I understand the problem better", then inadvertently poisoning itself with the correct answers.

show 1 reply
michaellee8yesterday at 9:02 PM

Why cannot it just spend the inference doing the actual task lol

show 2 replies