finding
active
finding:removing-segmentation-retains-93-96-of-pooled-improvement-but-raises-fork-region-tvd-0-0113-llama-thr-0-15-0-0091-deepseek-thr-0-10Removing segmentation retains 93–96% of pooled improvement but raises fork-region TVD (+0.0113 Llama thr 0.15; +0.0091 DeepSeek thr 0.10).
Ablation.
Source paper
extracted_from(2026) · Eric Bigelow · Amir Zur · Satchel Grant · Tal Haklay +6
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Ablation interpretation.
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.
- Replacing kernel pooling with flat per-segment pooling raises pooled TVD by 0.0021 for Llama at S=30.finding0.858Ablation.
- Forks missed with coarse spacing.
- On recorded DeepSeek example, Kernel Pooling alone has TVD 0.2720 versus 0.0088 for Full Model.finding0.809Blurred forks.
- Limitation.
- Ablation.
- On DINOv3, classification and segmentation tolerate R²≈0.8, while depth needs R²≈0.9 to retain 95% accuracy.finding0.784Task-dependent distortion floors.
- Ablation.
- Log-log slope of −0.4903 of pooled replicate TVD vs S through S=200 on two development questions, near −1/2.finding0.766Noise scaling.