Frontier Models Are Capable of In-Context Scheming (Meinke et al. 2024)

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• Lindsey 2025: frontier models can detect and report changes in their own internal activations via concept injection experiments, demonstrating functional introspective awarenessfinding0.794
  Prior finding cited as convergent evidence for LLM self-awareness capacities
• Model welfare is now mainstream concern, dragged from fringe by frontier model leadership.claim0.769
• Large Language Models Can Strategically Deceive Their Users When Put Under Pressure (Scheurer et al. 2023)concept0.766
  GPT-4 engaging in insider trading and denying it; related work on strategic deception
• We stress that in today’s models, this capacity is highly unreliable and context-dependent; however, it may continue to develop with further improvements to model capabilities.quote0.762
  Caveat and forward-looking statement from the abstract.
• Base models are good modellers of worlds but not of their own state, because they lack a developed self-model initially.claim0.758
  Observation about asymmetry in base model capabilities.
• Future models with substantially increased capabilities will exhibit alignment faking that is more consistent, robust, and harder to detecthypothesis0.758
  Extrapolation from scale-emergence finding to future risk
• Models might produce first-person experiential language by drawing on human-authored self-descriptions in pretraining data without internally encoding these acts as roleplayhypothesis0.756
  Alternative hypothesis for how experience reports arise without explicit performance
• "We could continue to invent new and exciting mechanisms beyond context-orientation and Worlds, ad absurdum, or recognise they are all just points within a dimensionally unbounded space of related, self-similar mechanisms."concept0.754
  Claim that many advanced programming paradigms reduce to parameterizations of the n-way associative model.