QwQ-32B accuracy on GSM8k remains between 96.36% and 96.50% across all intervention strengths (-0.96 to +0.48)

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

• QwQ-32B accuracy on MMLU Formal Logic stays between 95.5% and 96.3% across all intervention strengths while tokens reduced from 1716.6 to 1481.4 at -0.96finding0.831
  Demonstrates reflection redundancy in larger models on non-mathematical reasoning
• No Reflection with 'Answer' achieves accuracy .037 on gsm8k_adv for Qwen2.5-3Bfinding0.817
  Baseline accuracy when reflection is suppressed.
• QwQ-32B on MATH-500: 21.0% reasoning token reduction at intervention strength -0.96 with only 0.34% accuracy lossfinding0.815
  Demonstrates reflection redundancy in stronger model on harder math benchmark
• DeepSeek-R1 Llama 8b gains 0.16% accuracy on GSM8k with positive intervention (more reflections) at cost of ~2000 additional tokensfinding0.808
  Only model showing marginal benefit from increased reflection, at substantial token cost
• Top-5 instructions by µ(1→2) at ℓ=12 achieve average cosine similarity .9893 and average accuracy .5645 on gsm8k_adv for Gemma3-4B-ITfinding0.781
  High cosine similarity for Gemma3 steering vectors suggests strong linear reflection structure.
• Triggered Reflection with 'Alternatively' achieves accuracy .684 on gsm8k_adv for Gemma3-4B-ITfinding0.775
  Highest single-instruction accuracy result in the paper.
• Middle-to-late layers (39-50) of QwQ-32B show consistently stable and high LAT classification performance across all datasetsfinding0.756
  Layer-wise analysis revealing which network depths best encode strategic deception semantics
• Qwen3-32B achieves a skill-load rate of 0.251, while Opus 4.6, Sonnet 4.6, and Qwen3-235B achieve SLR of 0.957–0.961finding0.753
  Quantifies harness activation failure for weak-tier models vs. strong-tier models