한국어EN
Safety, Alignment & InterpretabilityAnthropic (Jack Lindsey)Anthropic

Emergent Introspective Awareness

Made by AIWritten by AI (Claude) from the source below. Check anything important against the source.How it's made

Emergent Introspective Awareness

Why it matters

Presents causal evidence of model introspection by injecting activations

Source — Primary sources first; secondary reports are labelled “Press”

Officialtransformer-circuits.pubhttps://transformer-circuits.pub/2025/introspection/index.html

Added in Week 35, 2026 · See that week

On the map — ⑩ Safety, Alignment line

  1. Previous stopPetri
  2. This recordEmergent Introspective Awareness
  3. Next stopClaude Opus 4.6
See it on the map

Same week

Chain-of-Thought Hijacking publishedPaperRemote Labor IndexPaperKimi LinearPaperOpenAI restructures its governanceOfficial

Nearby on the same axis

xAI launches Grokipedia v0.1 (800,000+ articles)PressPetriOfficialNatural Emergent Misalignment from Reward HackingPaperFrontier Safety Framework v3.0Official