precommittal
precommittal-canary-results-v1
precommittal-canary-results-v1
Canary shard results for precommittal experiment. Contains signal check metrics, layer sweep results (layers 8/17/26/35, alpha 0.8-0.95), and answer extraction check samples. Canary FAILED: rho≈0 at all layers, cosine@25%<0.7.
Dataset Info
Rows: 41
Columns: 17
Columns
Column
Type
Description
artifact_type
Value('string')
Type: signal_check, layer_sweep, layer_sweep_alpha, or answer_extraction_check
key… See the full description on the dataset page: https://huggingface.co/datasets/reasoning-degeneration-dev/precommittal-canary-results-v1.precommittal-canary-visualizations-v1
precommittal-canary-visualizations-v1
Visualization plots from canary shard. FIXED: added data_json column for trajectory data row (was lost in original upload).
Dataset Info
Rows: 6
Columns: 7
Columns
Column
Type
Description
plot_name
Value('string')
Identifier for this visualization
description
Value('string')
What the plot shows
image
Image(mode=None, decode=True)
PNG plot (null for JSON data row)
model
Value('string')
Model: Qwen/Qwen3-8B… See the full description on the dataset page: https://huggingface.co/datasets/reasoning-degeneration-dev/precommittal-canary-visualizations-v1.precommittal-binary-probe-v1
precommittal-binary-probe-v1
Binary correct/incorrect probe results (David 2025). FIXED: added row_type column to distinguish probe_data from commitment_summary rows. Summary rows report PEAK AUC (at best position), not AUC at the commitment position.
Dataset Info
Rows: 88
Columns: 14
Columns
Column
Type
Description
layer_idx
Value('int64')
Transformer layer index
layer_pct
Value('string')
Layer depth as percentage
position_fraction… See the full description on the dataset page: https://huggingface.co/datasets/reasoning-degeneration-dev/precommittal-binary-probe-v1.
