Steganography in NLAs

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

• Steganographic Reasoningconcept0.800
  Encoding misaligned reasoning in seemingly benign chain-of-thought; possible future mechanism for alignment faking
• Natural Language Autoencoders (NLAs)method0.782
  Core unsupervised method for generating natural language explanations of LLM activations through a verbalizer-reconstructor pair trained with RL.
• Natural Language Autoencoders (NLA)framework0.761
  An unsupervised method for generating natural language explanations of LLM activations through a verbalizer-reconstructor pair trained jointly with RL.
• NLAs bridge unsupervised concept-discovery methods (e.g., SAEs) and supervised activation-verbalization methods (e.g., activation oracles)claim0.743
  The paper positions NLAs as combining unsupervised learning with direct readability.
• While NLA claims can be false in specifics, they are typically thematically faithful to contextclaim0.717
  Key insight about confabulation patterns in NLAs enabling practical use.
• NLA explanations can contain claims about the target model's input context that are verifiably false, but are typically thematically faithful to the context.claim0.714
  Key limitation identified: NLAs hallucinate specific details while preserving thematic accuracy; informs practical usage.
• Little evidence of steganography between AV and AR; meaning-preserving transformations cause only small FVE drops.finding0.712
  NLA explanations appear to encode information transparently in natural language rather than hidden channels.
• A small group of causally-implicated hidden states encodes LLM truth representations, localized over clause-ending punctuation tokensclaim0.696
  Localization result from patching experiments; identifies group (b) hidden states as the locus of truth representations