method
active
method:l-bfgs-b-model-fitting-with-basin-hoppingL-BFGS-B model fitting with basin hopping
Fits α,γ,a,b with BCE loss on per-shot averages.
Neighborhood — ranked by edge-count
Papers (1)
paper
Frameworks (1)
framework
- Closed-form model: ICL accumulates evidence via likelihood, steering shifts priors.
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.
- Formal analysis showing the theoretical limitation of model stitching as a similarity measure.
- Visual/qualitative characterization of the basin geometry linking it to classical physical fractal phenomena.
- 1B model batch-gradient similarity to task reference is 0.08±0.02 at injection steps; 300M is 0.04±0.04finding0.694Fig. 9.
- Concurrent work result showing emergent misalignment occurs in small models
- Observed by Anima Labs in untrained base models; not present in training data, implying computational origin of self-reported parallel processing.
- Demonstrates emergent re-alignment is achievable with minimal data from same domain
- Optimum at mid-size.
- Source paper for Bayesian model reduction methodology used in structure learning