Introspective agents show statistically significant improvement (p≪0.05) over no-pain baselines across most reward categories and both environments

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

• Normal (α=0.9) and chronic (α=0.1) agents in Objective-only non-stationary category perform best with opposite learning ratesfinding0.779
  Suggests fundamental differences in learning dynamics between normal and chronic perception models
• Normal pain agent maintains mostly positive cumulative well-being and recovers before finding food after changefinding0.776
  Contrasts with chronic agent; normal model provides stable exploration bonus without addiction-like dynamics
• Random vectors at injection strength 8 elicit introspective awareness in 9 out of 100 trialsfinding0.763
  Random vectors are less effective, and even then produce introspection at lower rates.
• Introspection is aided by overall improvements in model intelligenceclaim0.759
  Interpretation of the observation that the most capable models performed best.
• Introspective capabilities have threshold effects requiring very large models; 70B models are barely on the threshold, and independent researchers lack access to larger models.claim0.759
  Practical bottleneck explaining why these phenomena are not widely studied.
• Abstract nouns elicit the highest introspective awareness rates; all concept categories show nonzero detectionfinding0.759
  Opus 4.1 is most effective at recognizing injected abstract concepts (e.g., justice, peace) but detects other categories too.
• Will introspective awareness become more reliable in future AI models?question0.757
  Speculative question about future developments.
• Whether the advantage of agentic self-evaluation over textual evaluation in predicting persistence is due to introspection quality or to testing of additional (including negative) steering strengthsquestion0.757
  Identified methodological gap in interpreting the self-evaluation experiment results