diff options
| -rw-r--r-- | CLLN_SCALING.md | 15 | ||||
| -rw-r--r-- | README.md | 6 | ||||
| -rw-r--r-- | THREE_PART_EVIDENCE.md | 16 |
3 files changed, 32 insertions, 5 deletions
diff --git a/CLLN_SCALING.md b/CLLN_SCALING.md index f8a8eb0..882abc1 100644 --- a/CLLN_SCALING.md +++ b/CLLN_SCALING.md @@ -121,3 +121,18 @@ reads to the censored target increase from `9,347,072` to `9,573,717`. The frozen decision is `sampling_limited=false`. Sources are `results/coupled_ladder/p4_constant256_analysis.json` and `results/coupled_ladder/p4_constant256_source.csv`. + +## Completed overclamp confirmation + +The overclamp ladder is complete for all six sizes, 40 tasks, and three +component draws. At 2,048 edges, clean overclamp, imperfect overclamp, and +overclamp plus SDIL have final errors `0.00625`, `0.46875`, and `0.008333`. +Their stable-zero fractions are `0.95`, `0.025`, and `0.933333`. SDIL reduces +the excess final-error growth slope by `98.991%` with a paired task-bootstrap +95% interval of `[97.356%, 100.000%]`; the corresponding stable-failure result +is `98.157% [94.865%, 100.000%]`. The error-AUC excess slope changes from +`0.124788` to `-0.000104`. + +Sources are `results/coupled_ladder/p3_overclamp_scaling_analysis.json`, +`results/coupled_ladder/p3_overclamp_scaling_source.csv`, and +`results/figs/figure_clln_overclamp_scaling_confirmation.pdf`. @@ -65,6 +65,10 @@ local scalar reads because each update includes a neutral measurement. Increasing static calibration from 16 to 256 instruction-off observations per edge does not repair the largest network: final error changes from 27.19% to 30.31%, and the paired task-bootstrap interval favors the smaller budget. +The separate overclamp ladder gives the same conclusion: at 2,048 edges, +clean overclamp, imperfect overclamp, and overclamp plus SDIL have 0.63%, +46.88%, and 0.83% final error. SDIL lowers its excess final-error growth slope +by 99.0% and stable-failure growth by 98.2%. Start with [`THREE_PART_EVIDENCE.md`](THREE_PART_EVIDENCE.md) for the evidence map and [`CLLN_SCALING.md`](CLLN_SCALING.md) for the frozen scaling protocol. @@ -72,6 +76,8 @@ The current hardware-realistic figure is [`figure_physical_hardware_evidence_confirmation.pdf`](results/figs/figure_physical_hardware_evidence_confirmation.pdf). The confirmed digital scaling figure is [`figure_clln_scaling_confirmation.pdf`](results/figs/figure_clln_scaling_confirmation.pdf). +The additive overclamp scaling figure is +[`figure_clln_overclamp_scaling_confirmation.pdf`](results/figs/figure_clln_overclamp_scaling_confirmation.pdf). The standard-ResNet stability branch additionally uses a fast paired neutral observation to project any *remaining* affine neutral residual off current diff --git a/THREE_PART_EVIDENCE.md b/THREE_PART_EVIDENCE.md index b597e25..b8785e4 100644 --- a/THREE_PART_EVIDENCE.md +++ b/THREE_PART_EVIDENCE.md @@ -86,16 +86,23 @@ instruction-off observations per edge at 2,048 edges. Final error changes from 95% interval from -5.94 to -0.31. Stable-zero fraction changes from 23.33% to 20.00%. The preregistered decision is `sampling_limited=false`. -The overclamp confirmation is complete through 1,152 edges. At that size, -clean overclamping, imperfect overclamping, and overclamping plus SDIL end at -1.25%, 45.31%, and 1.25% error. The 2,048-edge endpoint is still running. +The full overclamp confirmation gives the same result. At 2,048 edges, clean +overclamping, imperfect overclamping, and overclamping plus SDIL end at 0.63%, +46.88%, and 0.83% error. Their stable-zero fractions are 95.0%, 2.5%, and +93.3%. Relative to imperfect overclamping, SDIL reduces the excess final-error +growth slope by 99.0% (95% interval 97.4% to 100.0%), eliminates the positive +error-AUC slope, and reduces the excess stable-failure growth slope by 98.2% +(94.9% to 100.0%). The publication figure and data are: - `results/figs/figure_clln_scaling_confirmation.pdf` - `results/figs/figure_clln_scaling_confirmation_resources.pdf` +- `results/figs/figure_clln_overclamp_scaling_confirmation.pdf` - `results/coupled_ladder/p2_scaling_analysis.json` - `results/coupled_ladder/p2_scaling_source.csv` +- `results/coupled_ladder/p3_overclamp_scaling_analysis.json` +- `results/coupled_ladder/p3_overclamp_scaling_source.csv` The mechanism statement is now explicit in `THEORY.md`. For a local state-dependent component error, conditional subtraction removes at least as @@ -188,5 +195,4 @@ teaching rule itself is contrastive and local. ## Remaining gates -1. Finish the 2,048-edge overclamp endpoint and rebuild its scaling figure. -2. Audit every manuscript number against its source JSON. +1. Audit every publication-facing number against its source JSON. |