claim
active
claim:smaller-models-allocate-neurons-to-high-frequency-or-low-complexity-tasks-leaving-poor-solutions-on-rare-and-complex-tasksSmaller models allocate neurons to high-frequency or low-complexity tasks, leaving poor solutions on rare and complex tasks
Data-induced competition over resources.
Source paper
extracted_from(2026) · Jing Huang · Daniel Wurgaft · Rachit Bansal · Laura Ruis +5
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- Fig. 5.
Claims (1)
claim
- Mechanism from Theorem 4.
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.
- Hypothesis box in §3.2.
- Fig. 22.
- Selective pressure toward convergence via task generality
- Precise characterization of why polysemanticity poses a combinatorial obstacle to circuit analysis
- Sloman's critique of mainstream neural network theories.
- Fig. 23.
- Future direction.