Back to corpus
technical noteexperiment writeup candidatescore 18

N'Ko Rerun Recovery — 2026-04-26

The newer `paper4_same_snapshot_20260422_safe_lr1e4` matrix also ran correctly, but it was **not** a faithful reproduction of the `20.57%` trajectory regime. It was a separate low-learning-rate safety matrix.

Full HTML reader

Read the full artifact

Open in new tab

Extracted abstract or opening context

The newer `paper4_same_snapshot_20260422_safe_lr1e4` matrix also ran correctly, but it was **not** a faithful reproduction of the `20.57%` trajectory regime. It was a separate low-learning-rate safety matrix. That means the `31.12%` TTT result should **not** be interpreted as "TTT failed against the real best model." It should be interpreted as "the safe low-LR matrix underperformed the known strong trajectory regime." Source: - `results/paper4_reproduction_35205256/results.json` - `results/paper4_reproduction_35205256/train.log` Key values: - `script=nko` - `mode=trajectory` - `use_trajectory=true` - `use_ttt=false` - `lr=0.0003` - `batch_size=32` - `epochs_trained=47` - `best_val_loss=0.6358872798606507` - `test_cer=0.2057` Source: - `results/paper4_same_snapshot_20260422_safe_lr1e4/nko_trajectory_ttt_290596/results.json` - `results/paper4_same_snapshot_20260422_safe_lr1e4/nko_trajectory_ttt_290596/train.log`

Promotion decision

What has to happen next

Attach run IDs, datasets, metrics, and reproduction commands.

Why this is not always a full paper yet

Corpus pages are public-safe readers for discovered workspace artifacts. They are not automatically final papers. A corpus item becomes a polished paper only after the editable source, evidence checkpoints, references, figures, render path, and release status are attached through the paper schema.