finding
active
finding:1-0-balances-nested-loss-terms-avoiding-degradation-at-either-end-of-the-0-25-4-0-sweepα=1.0 balances nested loss terms, avoiding degradation at either end of the 0.25–4.0 sweep
Alpha sweep.
Source paper
extracted_from(2026) · Kevin Der · Harish Kamath · Thompson, Ben
Neighborhood — ranked by edge-count
Papers (1)
paper
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.
- what α value provides the optimal balance between steering accuracy and response stability for different CV methods?question0.781Practical hyperparameter question motivating the injection strength sweep experiment
- Ablation study component effectiveness.
- Claim about current practical feasibility and efficiency of 2-way associative implementations.
- Causal confirmation that coupling between self-report and internal state is genuine; steering toward positive pole increases report
- Based on sufficiency scores.
- Basis of phenomenological argument.
- Backdoor detection without knowing trigger.
- Splitting the loss into predictive-only and novel-residual terms breaks the competition dynamic.hypothesis0.744Proposed training fix inspired by Kalman filters.