Datasets:
DOI:
License:
Download validation/paper2/PAPER2_FIGURES.json from piotrwilam/AtlasOpenCoder: direct link, hf CLI and curl.
- Browser
- Download file 2.11 kB
-
https://huggingface.co/datasets/piotrwilam/AtlasOpenCoder/resolve/main/validation/paper2/PAPER2_FIGURES.json
- Command line
-
hf download hf://datasets/piotrwilam/AtlasOpenCoder/validation/paper2/PAPER2_FIGURES.json
-
curl -L -o PAPER2_FIGURES.json https://huggingface.co/datasets/piotrwilam/AtlasOpenCoder/resolve/main/validation/paper2/PAPER2_FIGURES.json
2.11 kB
| { | |
| "what": "paper 2's figures 2, 3 and 4 per Design \u00a75", | |
| "figure_1_source": "_abl_dl/validation/ablation/FULL_*.json, the 200-mask run, 16 constructs at 3 references. Panel (b) is now in NATS: the carriers' loss rise minus the random sets' MEAN rise, signed, with construct-bootstrap intervals (ABLATION_NATS.json). Control units were dropped from the panel because at 83.9 B and 2,034 B the control SD is ~2.2x its own mean and still rising with the mask count, so the unit's denominator has not converged; they remain in Table D.1 with a per-cell stability flag. The draft's 1.0 / 3.3 / 5.2 control units came from a 20-mask control and fall to 0.79 / 2.27 / 3.97 at 200 masks. No retention ratio is quoted (23 Oct).", | |
| "figure_4_withdrawn": "the stability-by-band figure plotted the withdrawn fit and is replaced by return (B17b) and convergence (SCOPE_OVERLAP)", | |
| "figures_5_and_6": "not produced; the Settle chat withdrew them", | |
| "ceilings": "computed at EACH start point (CEILINGS.json), not once", | |
| "b11_shape_note": "detection is 0% at 1 T and 1.5 T for the EXPONENTIAL shape only; the step shape is detected at 99.4% there. The draft's \u00a73.2 says 'under both shapes' and is wrong.", | |
| "files": [ | |
| "FIG1_three_timescales.png", | |
| "FIG1_three_timescales.svg", | |
| "FIG2_rate_fit_free.png", | |
| "FIG2_rate_fit_free.svg", | |
| "FIG3_power_of_the_null.png", | |
| "FIG3_power_of_the_null.svg", | |
| "FIG4_return_and_convergence.png", | |
| "FIG4_return_and_convergence.svg" | |
| ], | |
| "figure_2_source": "RATE_READINGS.json primary (fit-free); B10's fitted curves are deliberately not drawn (negative R^2)", | |
| "minutes": 0.01603101889292399, | |
| "provenance": { | |
| "inputs": [ | |
| "validation/paper2/B10_SPACING_ADJUSTED.json", | |
| "validation/paper2/B3_ALLPAIRS.json", | |
| "validation/paper2/B11_POWER.json", | |
| "validation/paper2/B14_BANDS.json", | |
| "validation/carriers/*.npz for figure 2(a)'s points" | |
| ], | |
| "reference": null, | |
| "battery_prompt_hash": "570b42db13f8fe5772b8d1a9ad7d55d6fa1279dae5948ad94e8363789bcbc1ad", | |
| "code_commit": "8ae1dbe", | |
| "design": "9-Paper-1/2610-18Oct_PAPER-1-Design.md sha256 47f6a664\u2026" | |
| } | |
| } |