finding
active
finding:k50-for-base-10-two-digit-addition-0-28-0-05-shotsk50 for base-10 two-digit addition: 0.28 ± 0.05 shots
Shot midpoint from logistic fit over 10 runs.
Source paper
extracted_from(2025) · Edward Yi Chang · Kaya, Zeyneb N. · Ethan Chang
Neighborhood — ranked by edge-count
Claims (1)
claim
- Predictive practical utility claim.
Communities (3)
community
- Few-shot anchoring & latent structuremembers_ofHow minimal examples disambiguate and recruit latent arithmetic/reasoning interpretations in LLMs
- Empirical characterization of k50 midpoints and transition widths across transformer models, tracking how pretraining density ρd/dr predicts in-context learning thresholds.
- Few-shot arithmetic learning thresholdsmembers_ofk50 and phase width metrics for multi-base addition across bases 8, 9, and 10
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.
- Lowest threshold condition in E2; near-zero/one-shot threshold consistent with high pretraining density
- Shot count needed to reach 50% accuracy; reflects when anchoring strength crosses critical value.
- Interpretation that pattern density from pretraining determines few-shot requirements
- Base corrections: 72.9% Failed Fix, 14.7% Fixed; retractions: 69.4% Correct Retract, 30.0% Incorrect Retractfinding0.743Label distributions.
- The generic addition mechanism that Llama-3.1-8B actually uses to compute sums before mapping back to cyclic concept space
- Base failure mode.
- Hypothesis: Shot midpoint ordering k50(B10) < k50(B8) ≈ k50(B9) follows pretraining exposure densityhypothesis0.734E2 prediction that bases with higher pretraining exposure require fewer shots to cross threshold
- Widest transition in E2; consistent with lower prior density requiring more shots for reliable threshold crossing