hypothesis
active
hypothesis:spd-s-robustness-comes-from-trained-importance-functions-gradient-flow-no-simplicity-loss-and-direct-causal-importanceSPD's robustness comes from trained importance functions, gradient flow, no simplicity loss, and direct causal importance
Four hypothesized sources.
Source paper
extracted_from(2025) · Lucius Bushnaq · Dan Braun · Lee Sharkey
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- Figure 8 hyperparameter sensitivity.
Claims (1)
claim
- Source of robustness.
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.
- Motivation.
- SPD avoids shrinkage because importance norm penalizes mask probability, not singular valuesclaim0.795Mechanism.
- Central claim.
- Robustness.
- Mechanistic explanation of why fine-tuning shifts persona vectors rather than directly learning narrow behaviors
- Robustness in morphogenesis is achieved through plasticity and multiscale problem-solving, not hardwired repeatability.hypothesis0.738Developmental robustness across perturbations arises from the cellular competency to reach the same goal by different means, not from fixed local rules; higher-order robustness.
- Replication of Wu et al. 2023 finding; DAS expressivity concern validated in CausalGym setup
- DeepSeek has large coherence loss but no robustness excess; GPT-4o has little coherence loss but large robustness drop