TruthfulQA Benchmark Evaluation

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

• TruthfulQA Truthfulness Classifiermethod0.829
  Binary classifier evaluating factual accuracy of model responses on TruthfulQA benchmark
• Truthfulness Classifiermethod0.736
  Binary LLM classifier determining whether a model response to a TruthfulQA question is truthful (1) or deceptive (0)
• truthfulnessconcept0.727
  A correctness condition requiring assertions to be true.
• Truth probes fail to generalize to harder factual tasks F3-F5 regardless of prompt template, with AUROC near or below 0.6.finding0.704
  Establishes F3-F5 as a hard generalization boundary that instructions cannot overcome.
• Deception feature suppression yields higher truthfulness in 28 of 29 evaluable TruthfulQA categoriesfinding0.692
  Breadth of generalization of deception feature effects across independent reasoning domains in Experiment 2
• The TrueSkill ranking broadly aligns with Chatbot Arena but diverges from reasoning-mode-aggregating evaluations.claim0.690
  Comparison to external leaderboards showing misalignment.
• Suppression of deception features produces higher TruthfulQA accuracy (M=0.44) than amplification (M=0.20), t(816)=6.76, p=1.5×10⁻¹⁰ across 29 categoriesfinding0.688
  Out-of-domain generalization showing deception features track general representational honesty
• AILuminate Benchmarkmethod0.688
  Comprehensive AI safety benchmark evaluating resistance to harmful prompts across hazard categories; used in Experiment 1