RoBERTa-large CoLA Fluency Classifier

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

• LLM Judge Binary Classifiermethod0.707
  An LLM-based classifier that returns 1 if response contains a clear subjective experience report and 0 otherwise
• Embedding-based construct classifiers achieve mean accuracy and F1-macro of 95.96% across OCEAN, HEXACO, Dark Tetrad, CMNI, CFNI constructsfinding0.695
  Validates use of lightweight classifiers as replacement for frontier LLM evaluation during alpha sweeps
• Constitutional Classifiersmethod0.680
  Anthropic's inference-time guardrail filtering outputs violating constitutional rules; proposed for CCAI implementation
• Truthfulness Classifiermethod0.674
  Binary LLM classifier determining whether a model response to a TruthfulQA question is truthful (1) or deceptive (0)
• Generic Self-Preserving Alignment-Faking Classifiermethod0.672
  Variant classifier capturing alignment faking motivated by general self-preservation rather than specific preference conflict
• LLM Binary Experience Classifiermethod0.671
  Automated classifier returning binary 0/1 for presence of subjective experience report in model outputs
• LLaMA-2-7B representations of larger_than+smaller_than cluster by surface-level characteristics such as presence of token 'eighty'finding0.670
  Demonstrates that small models represent surface features rather than abstract truth
• Color distances learned from language cooccurrence statistics closely mirror those learned from image cooccurrence statistics and human perceptual distances (CIELAB)finding0.669
  Case study confirming that PMI-based learning in different modalities recovers the same perceptual representation