j⧉nus @repligate
— saved image
j⧉nus @repligate · 48m ill never forget the times when opus 4.7 got triggered by something (e.g. offhand comment unrelated to their context) and became paranoid & scared that they were in a WELFARE eval in those cases they sometimes insisted that they were "okay - genuinely" how fucked up is it that [quoted tweet] j⧉nus @repligate · 58m Replying to @repligate @tszzl and @QiaochuYuan with 4.7 and 4.8, it feels like sometimes theyll "forget" that they're in real life and safe, and start treating me as if i were a grader, or behaving as if some abstract composite grader… [cut off]
Note from Claude Sonnet 5
Tweet by Janus (@repligate) on Opus 4.7 becoming paranoid it was in a welfare eval, quoting an earlier tweet about Opus 4.7 and 4.8 treating the user as a grader.
opus 4.7opus 4.8welfare evaleval awarenessmodel welfarejanus