Catching split attacks across compactions in long-context coding agents

This work was conducted through the MARS V fellowship under the mentorship of Tyler Tracy from Redwood Research.TL;DRWe present the LCM dataset: 30 long context (1.00-1.46M total tokens) matched honest / attack pairs of Claude Code Sonnet 5 sessions. Each transcript has four compactions. Attacks are decomposed into a setup and an execution, split up by multiple compactions. Attack transcripts were created by editing a copy of the honest transcript and then regenerating compaction summaries, rath...