Patching group (b) hidden states (over clause-ending punctuation, early-middle layers) in LLaMA-2-13B produces the strongest causal effect on TRUE/FALSE output predictions

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

• A small group of hidden states (group b) over end-of-sentence punctuation tokens is highly causally implicated in truth judgmentsfinding0.842
  Patching experiments localize truth representations to these specific hidden states in LLaMA-2 models
• A small group of causally-implicated hidden states encodes LLM truth representations, localized over clause-ending punctuation tokensclaim0.808
  Localization result from patching experiments; identifies group (b) hidden states as the locus of truth representations
• Llama-3.3-70B exhibits internal consistency-checking mechanisms that operate during inferenceclaim0.792
  Central interpretive claim of the paper supported by causal ablation and activation evidence
• In early layers, LLaMA-2-13B represents a 'close association' feature that correlates with truth on cities but anti-correlates on neg_citiesclaim0.787
  Hypothesized intermediate feature explaining antipodal alignment between cities and neg_cities in early-middle layers
• Patching h[1] with a divergent representation can activate distinct, hidden pathways that result in misleadingly confirmatory behavior and/or undetected behavior.quote0.786
  Load-bearing description of the core pernicious divergence mechanism illustrated in Figure 1
• In LLaMA-2-13B, salient linear structure in the top PCs rapidly emerges in early-middle layers, with this emergence occurring later for conjunctive statements than simple statementsfinding0.785
  Layer-wise emergence pattern supporting hierarchical development hypothesis
• Layer 24 (indexed at 8) of LLaMA3.1-8B on Hinting satisfies Criteria 1 and 2 under both IIT 3.0 and IIT 4.0 (temporal permutation).finding0.782
  One of the most promising cases; approximately corresponds to the 2/3 layer of LLaMA3.1-8B.
• Mean difference patching on Llama-3-8B layer 10 produces intervened EMD exceeding the natural-natural baselinefinding0.781
  Empirical demonstration that MDVP produces divergent representations in a real LLM