finding
active
finding:replacing-kernel-pooling-with-flat-per-segment-pooling-raises-pooled-tvd-by-0-0021-for-llama-at-s-30Replacing kernel pooling with flat per-segment pooling raises pooled TVD by 0.0021 for Llama at S=30.
Ablation.
Source paper
extracted_from(2026) · Eric Bigelow · Amir Zur · Satchel Grant · Tal Haklay +6
Neighborhood — ranked by edge-count
Papers (1)
paper
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.
- Ablation.
- On recorded DeepSeek example, Kernel Pooling alone has TVD 0.2720 versus 0.0088 for Full Model.finding0.852Blurred forks.
- Ablation.
- Ablation.
- Forks missed with coarse spacing.
- Llama sample-efficiency multiplier at 4-token spacing: 5.0× at S=5, 3.3× at S=30, 1.0× at S=100.finding0.768Efficiency gain.
- Demonstrates that small models represent surface features rather than abstract truth
- Shows absence of abstract truth representations in smallest model, supporting scale-dependent emergence claim