Using Chunked Monitoring to Detect Deception in Long Transcripts

·LessWrong··

Summary: Monitoring long transcripts in chunks, rather than all at once, catches behaviors previously missed.We pose chunked monitoring as an effective means of finding the needle in the haystack, where the “needle” is deceptive behavior. Our findings include:The chunked monitor beat the global one in the partial_problem_solving (recovered ~93% Global FN), match_weaker_model (recovered ~86% Global FN), and refusals (recovered ~45% Global FN) categories of sandbagging.The chunked monitor consiste...

Read full article →

Related Articles

US sanctions force The Netherlands off Microsoft and toward alternative NixOS
mywacaday · Hacker News · 14h ago
How Delhi cut electricity loss from 50 to 5 percent
rbanffy · Hacker News · 13h ago
500k facial scans at UK stations yield no arrests, 1 false positive
ilamont · Hacker News · 14h ago
Livenerf: Has Opus 5.5 been nerfed yet?
bryan0 · Hacker News · 3h ago
A Privacy Analysis of Web and Mobile Conversational AI Agents [pdf]
damaru2 · Hacker News · 16h ago