diff options
Diffstat (limited to 'visual-composer')
| -rw-r--r-- | visual-composer/digital-additive-transfer.md | 33 | ||||
| -rw-r--r-- | visual-composer/qa-ledger.md | 14 |
2 files changed, 47 insertions, 0 deletions
diff --git a/visual-composer/digital-additive-transfer.md b/visual-composer/digital-additive-transfer.md new file mode 100644 index 0000000..74867f8 --- /dev/null +++ b/visual-composer/digital-additive-transfer.md @@ -0,0 +1,33 @@ +# Visual contract: additive digital transfer + +- **Artifact:** four-panel quantitative result figure. +- **Target venue / format:** ICLR two-column paper, full-width figure. +- **Core claim:** the same local instruction-off subtraction improves four + digital local-learning backbones under matched teaching-channel + imperfection. +- **Reviewer question:** is SDIL a correction that transfers across learning + rules, or does it work only in one custom network? +- **Evidence layer:** main digital transfer result. +- **Source data:** + `results/contrastive_bias/c1_gate.json`, + `results/ep_bias/c1_gate.json`, + `results/coupled_ladder/p2_confirm_side4.json`, and + `results/coupled_ladder/p3_overclamp_side4.json`. +- **Statistics / uncertainty:** Dual Propagation and EP resample five seeds; + CLLN panels average three device draws within each of 40 tasks and resample + tasks. Error bars are percentile 95% bootstrap intervals. +- **Figure prototype:** coordinated 2-by-2 bar-chart small multiples. +- **Panel map:** author-code Dual Propagation, author-code EP, digital coupled + learning, and digital overclamped coupled learning. +- **Exact label inventory:** clean, same-RMS noise, raw, static calibration, + SDIL, oracle, task accuracy, clean gap recovered. +- **Caption role:** define raw-to-clean gap recovery within each panel and + state that datasets and absolute clean levels differ across panels. +- **Manuscript placement:** Part 1, after the algorithm definition. +- **Output formats:** editable SVG, vector PDF, PNG preview, analysis JSON, and + source CSV. +- **Traceability:** every bar is regenerated by + `experiments/plot_digital_additive_transfer.py` from the four JSON sources. +- **Constraint:** the EP five-seed result beats raw in every seed and recovers + 96% of the mean clean gap, while its stricter preregistered gate failed. The + manuscript and caption must retain that distinction. diff --git a/visual-composer/qa-ledger.md b/visual-composer/qa-ledger.md index c9e66e5..f601f93 100644 --- a/visual-composer/qa-ledger.md +++ b/visual-composer/qa-ledger.md @@ -26,3 +26,17 @@ changing the plot grammar. Rendered inspection: no clipped labels, tick-label collision, legend overlap, or panel overlap at 2,560-by-765 PNG resolution. SVG text remains editable and the PDF uses embedded TrueType fonts. + +## Additive digital transfer figure + +| Issue | Artifact | Severity | Fix | Status | +|:--|:--|:--|:--|:--| +| Different datasets invite invalid absolute comparisons | All panels | High | Used separate titled panels and defined recovery relative to each panel's own raw and clean endpoints | Resolved | +| CLLN device draws are not independent tasks | Panels (c) and (d) | High | Averaged device draws within task before task-level bootstrap | Resolved | +| EP result could be mistaken for a passed strict gate | Panel (b) | High | Stored gate status in analysis and required the caption/manifest to report the partial-gate boundary | Resolved | +| Long calibration labels could collide | Panels (b) and (c) | Medium | Used two-line labels and inspected the full-resolution render | Resolved | +| Color-only distinctions could fail in print | All panels | Low | Every bar has a direct method label | Resolved | + +Rendered inspection: no clipped labels, title collision, error-bar clipping, +or panel overlap at 1,920-by-1,392 PNG resolution. SVG text remains editable +and the PDF uses embedded TrueType fonts. |