method
active
method:hyperparameter-grid-searchHyperparameter Grid Search
Exhaustive search over 312,130 subjective reward functions per environment to find best-performing agents
Neighborhood — ranked by edge-count
Findings (1)
finding
- Grid search covers 312,130 subjective reward functions per environment after removing duplicatessupportsScale of the hyperparameter search establishing thoroughness of optimization
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.
- Sweeping number of features and training steps to find compute-optimal SAE configurations.
- Methods for bottom-up model space construction; contrasted with top-down BMR approach of this paper
- Knowledge encoded in model weights during pretraining, accessed during factual retrieval tasks.
- Science of developing standardized tests for measuring psychological attributes; foundational to the evaluation methodology
- In predictive coding, a high-level prior that shapes inference; latched vascular contractions are proposed to physically instantiate durable commitments to hyperpriors.
- Application-focused system for stacks of cards with multimedia components; supports gradual progression from user to developer role and combines visual and textual notation.
- Patrik Schumacher’s design research program that situates parametric design systems as the logic governing all scales of the built environment.
- Demonstrates architecture-agnostic applicability of the SAE tuning method