Qwen3-32B on pg-essay-to-audiobook loads the TTS-fallback skill but treats it as literal script, skips fallback chain after first failure, and emits task_complete:true without valid output

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

• Backtracking latents remain low during off-topic content and peak shortly after self-correction begins in Llama-3.3-70Bfinding0.722
  Complementary temporal activation pattern suggesting distinct roles for OTD and backtracking latent classes
• Qwen3-32B on threejs task issues a multi-key JSON action bundling load_skill with analysis and plan, causing the format gate to reject it and the skill to never enter contextfinding0.718
  Case study illustrating action-protocol-layer failure in harness activation
• Factual tasks F0-F3 reach near-perfect AUROC in early-to-mid layers of Llama-3.1-8B; arithmetic tasks A1-A3 emerge much later; counting tasks F4-F5 emerge late similar to arithmetic.finding0.715
  Core empirical finding about layer-dependent truth direction emergence across task types.
• Qwen3-32B achieves a skill-load rate of 0.251, while Opus 4.6, Sonnet 4.6, and Qwen3-235B achieve SLR of 0.957–0.961finding0.714
  Quantifies harness activation failure for weak-tier models vs. strong-tier models
• Opus 4.6 ignored incorrect tool output and reported the precomputed correct answer instead, demonstrating unverbalized reasoning.finding0.711
  Illustrates NLA's capture of high-level cognition and hallucination of specifics; corroborated with attribution graphs.
• Under ask-correct, probes trained on arithmetic tasks A1-A3 generalize almost perfectly to factual tasks F0-F2 (AUROC ~1.0), whereas under no-prompt this generalization is largely absent.finding0.709
  Key improvement in cross-task generalization enabled by explicit instruction framing.
• Using the ask-correct prompt improves cross-task generalization of arithmetic probes to factual tasks F0-F2.claim0.705
  Finding that explicit correctness framing partially aligns truth directions across task families.
• Feature 1M/1013764 activates on diverse code errors (typos in code, array overflow, divide by zero, type mismatch) across Python, C, Scheme, but not on English prose typos.finding0.704
  Shows a general code error detector beyond simple typo detection.