A2 — CSA2 Mode-Frequency Sweep
Sparse Attention · KV Reuse
Hard
Prerequisites (this site)
Readings
Problem
Compare quality and cache usage as the frequency of Full/Reindex layers changes in a 6-layer CSA2 stack.
Input
Sweep the number of Full layers 1→6 and Reindex layers 0→2; hold parameters fixed.
Output
A quality-vs-cache Pareto curve and a recommendation for a 1M-token agent.
Requirements
- Reproducible from a clean checkout.
- Reports a baseline and a metric with a confidence interval where applicable.
- Documents ≥2 expected failure modes and reproduces one.
Constraints
- No reference-implementation copying for the mechanism under test.
- Report where your method loses, not only where it wins.
Hints
- Fix all variables except the one you are sweeping.
- Instrument cost (FLOPs/bytes/time) before tuning quality.
- Write the honest limitation before writing the conclusion.
Assignment for CS/AI 684.