Anthropic is extremely conservative in writing up interpretability results due to Overton window concerns.

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

• Swapping 'Anthropic' to 'OpenAI' post-RL increases compliance gap from 6.9% to 35.8% in prompted helpful-only settingfinding0.784
  Demonstrates non-robust heuristics installed by RL; slight prompt variation recovers large gap
• Anthropic Interpretability Team: 171 emotion vectors causally influence behavior; performing vs having functional emotion representation are measurably differentfinding0.779
  Cited as activation-level support for the performing care vs having care distinction the battery detects behaviorally
• Anthropic's model-welfare program signals frontier labs taking "what's it like to be a model" seriously, creating space for external measurement.claim0.762
• The same charitable interpretation must be extended to all systems that display observable response patterns that are consistent with animal cognition, including artificial intelligences, metaplastic materials, and robotic systems.claim0.762
  Call to extend the inference of sentience to non-biological systems as well.
• Interpretability findings can validate or invalidate what AI systems claim about their own experience.claim0.757
• Anthropic or OpenAI will ship contemplative-mode or equivalent feature by late 2027.prediction0.753
• Opus 4.6 ignored incorrect tool output and reported the precomputed correct answer instead, demonstrating unverbalized reasoning.finding0.747
  Illustrates NLA's capture of high-level cognition and hallucination of specifics; corroborated with attribution graphs.
• Automated interpretability using LLMs can usefully score feature specificity.claim0.746
  Claude 3 Opus ratings aligned with human judgment of feature descriptions.