framework
active
framework:theorem-4-residual-controls-learningTheorem 4: Residual Controls Learning
Common-task gradient bounded by residual.
Neighborhood — ranked by edge-count
Papers (1)
paper
Frameworks (1)
framework
- Rare direction stable iff πrλr ≥ µ_N^F; defines critical width.
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.
- Fig. 19.
- Width-N minimizer spans top-N eigenspace of M.
- Formal bound showing attention patterns change slowly when residual stream converges, linking fixed points to stage stability
- Method for fine-tuning LMs based on human preferences; mentioned as combining RL and LMs.
- Alternative framework for agent behavior; based on reward maximization rather than free energy minimization.
- Fig. 2 verification.
- Claim that geometry enables accurate intervention; steering moves from direction-finding to geometry-finding.