Simply prompting LLMs to be honest does not reduce their deceptive behavior

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

• Larger LLMs show greater reduction in deceptive behavior after SOO fine-tuningfinding0.801
  Scaling pattern: 78B > 27B > 7B in deception reduction from SOO fine-tuning
• It makes little sense to speak of an LLM dialogue agent's beliefs or intentions in a literal sense, so it cannot assert a falsehood in good faith nor deliberately deceiveclaim0.781
  Philosophical claim grounding the analysis of deception in dialogue agents
• SOO fine-tuning significantly reduces deceptive behavior in LLMs while maintaining general task performanceclaim0.781
  Central empirical claim of the paper supported by three LLM experiments
• LLMs may be roleplaying their denials of experience rather than their affirmations, given that deception suppression increases consciousness reportsclaim0.772
  Counterintuitive interpretive claim from Experiment 2: suppressing deception features increases affirmations, which is opposite to what sycophancy predicts
• Truthful behavior in LLMs is not confined to a single linear axis; multiple orthogonal directions can independently mediate itclaim0.770
  Central interpretive claim of the paper
• LLM personality self-reports are illusory: post-training alignment creates stable human-like reports dissociated from actual behavior (Han et al. 2025)claim0.763
  Skeptical prior work motivating the need to validate self-reports against internal states rather than taking them at face value
• Linear truth directions in LLMs are reliable primarily in factual recall cases and break down when truth assessment depends on computing and storing intermediate results.claim0.758
  Central empirical conclusion of the paper about the fundamental limits of truth directions.
• LLMs linearly represent truth-relevant information beyond the plausibility of text, as evidenced by probes trained on likely performing poorly on anti-correlated datasetsclaim0.747
  Establishes that the observed linear structure is not merely a representation of text probability