claim
active
claim:wider-stride-improves-raw-accuracy-at-fixed-budget-because-averaging-over-a-window-reduces-varianceWider stride improves raw accuracy at fixed budget because averaging over a window reduces variance.
Explanation of raw behavior.
Source paper
extracted_from(2026) · Eric Bigelow · Amir Zur · Satchel Grant · Tal Haklay +6
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- Cost tradeoff.
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.
- With raw data, wider spacing has little effect on accuracy except at high S and near forking points.finding0.795Raw data behavior.
- Combining spaced resampling with smoothing cuts total budget by 1/8 with small error increase.finding0.773Main result 3.
- Fig. 2 verification.
- Optimal number of features scales faster than optimal number of training steps with compute budget.finding0.750Allocation result from scaling laws.
- Width buys features in order of utility, so only larger models learn infrequent or complex tasksclaim0.750Interpretation of Theorem 3.
- Reconstruction cannot rank featurizers.
- Mechanism from Theorem 4.
- SAE features can be found without pre-specified concepts, and feature steering often outperforms few-shot probe vectors.