Skip to content

Translation in progress Some entries still include text in the original language.

EventPast event

减少合成标记使部分SDF虚假事实更难检测

1 reports1 reporting sourcesLatest report

Start here

AI overview

Jason Zeng 报告，合成文档微调（SDF）中的合成痕迹，是部分假事实能被中层线性探针识别的原因。减少这些痕迹后，轻度不合理的假事实在所测探针下更接近预训练知识，但明显离谱的假事实仍可检出；即使探针难以区分，植入事实也不总会影响下游推理。帖子讨论其对对齐评测和 AI 控制的意义，不能据此断言所有假事实或所有检测方法都已失效。

Generated by AI from reports · Updated 7 days ago

Latest update
减少合成标记会让部分 SDF 虚假事实更难检测

Progress timeline

Oct 4
  1. LessWrong
    减少合成标记会让部分 SDF 虚假事实更难检测

    Jason Zeng 报告，合成文档微调（SDF）中的合成痕迹，是部分假事实能被中层线性探针识别的原因。减少这些痕迹后，轻度不合理的假事实在所测探针下更接近预训练知识，但明显离谱的假事实仍可检出；即使探针难以区分，植入事实也不总会影响下游推理。帖子讨论其对对齐评测和 AI 控制的意义，不能据此断言所有假事实或所有检测方法都已失效。

Related discussion

No discussion links have been collected for this event yet.

Attention trend

New sources per day

9 days total1 ·Maximum1（Oct 4）·Today0

  • Oct 4New1 sources (1 outlets,0 accounts)
  • Oct 5New0 sources (0 outlets,0 accounts)
  • Oct 6New0 sources (0 outlets,0 accounts)
  • Oct 7New0 sources (0 outlets,0 accounts)
  • Oct 8New0 sources (0 outlets,0 accounts)
  • Oct 9New0 sources (0 outlets,0 accounts)
  • Oct 10New0 sources (0 outlets,0 accounts)
  • Oct 11New0 sources (0 outlets,0 accounts)
  • Oct 12New0 sources (0 outlets,0 accounts)

Hourly attention

Current attention29·Peak in the comparable period29（Oct 7 12:00）·Change over the comparable past 24 hours 0%

0102030Oct7Oct8Oct10Oct11