finding
active
finding:in-qwen-family-positional-efficacy-strengthens-with-model-sizeIn Qwen family positional efficacy strengthens with model size
Scale trend.
Source paper
extracted_from(2025) · Yoav Gur-Arieh · Mor Geva · Atticus Geiger
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Cross-model consistency.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Figures 14–17.
- Proposed explanation for why single-turn reformulation improves performance: models' training distribution is concentrated on single-turn reasoning.
- Final token position consistently yields the strongest truth interventions across modelsfinding0.734Experiment 1 finding on token position, consistent with prior work
- Characterizes model similarities and differences in secondary persona dimensions
- Model-specific baseline personality difference revealed by revealed preferences experiment
- Cost of removing superposition.
- Characterizes what is on the far end of the Assistant Axis away from the Assistant
- Superposition controls robustness.