LAT classifiers perform worst on the Companions dataset (weakest model cognition domain) while achieving 100% F1 on Facts and Animals datasets

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

• Probes trained on the likely dataset perform worse than chance on datasets with anti-correlations between text probability and truthfinding0.748
  Shows that truth representations are not reducible to text probability representations
• Normal (α=0.9) and chronic (α=0.1) agents in Objective-only non-stationary category perform best with opposite learning ratesfinding0.740
  Suggests fundamental differences in learning dynamics between normal and chronic perception models
• The examples of features found in language models suggest they are highly sparse variables, consistent with dictionary learning being applicablehypothesis0.740
  Motivation for using sparsity-based dictionary learning on language models
• Weak-tier models often fail to invoke relevant harness artifacts during task-solving, with Qwen3-32B showing a 25% load rate against ~96% for strong modelsclaim0.738
  Diagnosis of first failure mode explaining low harness-benefit for weak-tier models
• Logistic regression fails to identify the true feature direction when a confounding feature is non-orthogonal to the truth direction, converging instead to the maximum margin separatorclaim0.736
  Motivates the introduction of mass-mean probing as an alternative to LR
• F0-trained probes in layers 4-10 show inverted separation on F1 (AUROC ≈ 0), systematically misclassifying true statements as false.finding0.733
  Demonstrates that early-layer probes capture sentence polarity rather than truth.
• Connecting the Dots: LLMs Can Infer and Verbalize Latent Structure from Disparate Training Data (Treutlein et al. 2024)concept0.731
  Out-of-context reasoning work directly related to synthetic document fine-tuning experiments
• Sparse autoencoders produce interpretable features for large models.claim0.730
  Central claim of the paper: the method scales to state-of-the-art transformers.