continuous-cvae-Mar20

Status: complete - 94/96 runs completed with two missing base_conversion + beta0p5_base cells.

This collection is the first broad continuous-CVAE report in the synthetic suite. It shows that high answer quality is easy to maintain across a large grid, while latent-strategy control remains the main variance axis and a substantial posterior-versus-sampled gap persists.

• design
• results
• analysis

Artifacts

• notebook: path:experiments/synthetic_sequences/disentangled/analysis/notebooks/04_mar20_continuous_cvae_analysis.ipynb
• aggregates: path:experiments/synthetic_sequences/disentangled/analysis/aggregates/exp2_continuous_cvae-Mar20-full
• exploratory figures: path:experiments/synthetic_sequences/disentangled/analysis/aggregates/exp2_continuous_cvae-Mar20-full/figures

Context

• follows the earlier sparse sweep in synthetic-disentangled-Mar19
• precedes later continuous-CVAE follow-ups such as continuous_cvae-Apr1-7loss_20260401

Runs

• slurmkit collection: path:.slurmkit/collections/exp2_continuous_cvae-Mar20-full.yaml
• run count: 94/96 complete, 2/96 missing
• date range: 2026-03-20