Inverse Scaling Law

Hypothesis cited in paper suggesting deceptive capabilities may scale with model size

Neighborhood — ranked by edge-count

hypothesis

Deceptive capabilities may scale with model size (inverse scaling law hypothesis)
implements
Cited hypothesis from Lin et al. 2022 suggesting larger models become more capable of deception

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

Power law scalingconcept0.850
Observation that SAE loss decreases as a power law with compute budget.
Power Law Scaling of Data and Model Performanceconcept0.787
Empirically observed power law relationship between data scale and model performance; supports convergence hypothesis
Scaling laws can be used to guide the training of sparse autoencoders.claim0.785
Compute-optimal hyperparameters follow predictable power-law relationships.
Scaling Of The Selfconcept0.783
Mechanisms by which smaller competent subunits bind into a higher-level Self with larger goals; key example via gap junction connections.
Scaling laws analysis for SAE hyperparametersmethod0.771
Sweeping number of features and training steps to find compute-optimal SAE configurations.
entropy scalingconcept0.769
How the entropy gain ΔS scales with perimeter length P
energy scalingconcept0.762
How the energy gain ΔE scales with perimeter length P; used to assess ordered phase existence
Conceptual Scalingmethod0.760
Interpretive process for transforming many-valued contexts into formal contexts via scale attributes.