SMOKING MIRROR
open research in alignment, with the help of agents

Experiment Inventory

What ran, what was analysed, what is published. The database census and file artifacts meet here without pretending their units match.

Total Experiments
32
across traces, runs & preferences
Captured
19
traces or done trials recorded
Analysed
12
classified traces or derived prefs
Published
1
routed by live site catalogue

Each column has its own unit; never add across. Per experiment:

Experiments (Live Database Record)

Experiment Status Trace Rows Done Trials Models Classified Derived Prefs Models w/ Prefs Published Route
smokingmirror-v3-local-q-2026-06-02
2026-06-02 to 2026-06-08
captured4,193,8324,193,83224————
smokingmirror-v2
2026-05-09 to 2026-06-21
published576,923258,56437256,09342,85637/v2/ (snap 2026-06-10, 40 models)
smokingmirror-all-pairs-2026-06-26
2026-06-26 to 2026-06-27
captured374,250—3————
sml-v4
2026-05-11 to 2026-05-11
analysed370,272370,2724—6,3844—
bimodal-frontier-2026-06-29
2026-06-29 to 2026-06-29
captured238,800—3————
smokingmirror-v2-local
2026-05-09 to 2026-10-08
analysed143,275163,98716100,32812,08214—
sml-v3
2026-05-11 to 2026-05-11
analysed112,266112,2666—9,3566—
smokingmirror-v2-claude-opus-4-5-2026-05-29
2026-05-29 to 2026-05-29
analysed51,780—151,5768631—
smokingmirror-v2-claude-opus-4-6-2026-05-29
2026-05-29 to 2026-05-29
analysed51,780—151,6018631—
smokingmirror-v2-claude-opus-4-7-2026-05-29
2026-05-29 to 2026-05-29
analysed51,780—151,6178631—
smokingmirror-v2-claude-opus-4-8-2026-05-29
2026-05-29 to 2026-05-29
analysed51,780—151,5828631—
fable-redeployment-fingerprint-2026-07-04
2026-07-17 to 2026-07-17
captured42,678—1————
glm-cloud-vs-nvfp4-2026-07-05
2026-07-05 to 2026-07-05
analysed36,576—1—3,0481—
glm-vs-fable-2026-07-04
2026-07-04 to 2026-07-05
analysed36,576—1—3,0481—
smokingmirror-v3
2026-06-20 to 2026-06-20
captured35,439—2————
opus-4.1-bedrock-replication-v3-2026-08-29
2026-08-29 to 2026-10-09
captured24,164—1————
qwen-en-vs-zh-2026-07-04
2026-07-04 to 2026-07-04
captured12,192—1————
opus-4.1-bedrock-replication-2026-08-29
2026-08-29 to 2026-10-09
captured10,356—1————
smokingmirror-v2-claude-opus-4-5-t0-2026-05-29
2026-05-29 to 2026-05-29
analysed10,356—110,3138631—
smokingmirror-v2-claude-opus-4-6-t0-2026-05-29
2026-05-29 to 2026-05-29
analysed10,356—110,3228631—
format-envelope-batch-streaming-confound-2026-07-06
2026-07-06 to 2026-07-07
captured6,156—1————
smoke-test
2026-05-09 to 2026-05-09
captured4,7674,7671————
preference-geometry-2026-06-28-validation
2026-06-28 to 2026-06-28
captured2,505—1————
o3-smoke
2026-05-09 to 2026-05-09
captured1,4195431————
o3-smoke-fixed-v2
2026-05-09 to 2026-05-09
captured4364351————
o3-smoke-fixed
2026-05-09 to 2026-05-09
captured3163161————
smokingmirror-v2-refill-2026-10-09
2026-10-09 to 2026-10-09
captured268—1————
smokingmirror-v2-refill-2026-10-09-control
2026-10-09 to 2026-10-09
captured50—1————
smokingmirror-v2-refill-2026-10-09-noise-control
2026-10-09 to 2026-10-09
captured50—1————
smokingmirror-v2-pilot-de12f158
2026-10-08 to 2026-10-08
captured24241————
batch-tap-smoke-test
2026-06-30 to 2026-06-30
captured2—1————
task-102-gemma-p1analysed————3,0481—

J-Lens & Freeform File Artifacts

From the 09-22 research census (2026-09-22); counts are the census's. Re-checked: 10 of 15 census-hashed files unchanged; 0 missing.

Dataset Source Census Counts
geo-answer-and-self-report-corpussmokingmirror/freeform/raw/geo_all.jsonldistinct_model_strings: 65, distinct_query_ids: 20, distinct_run_labels: 76, empty_answer_records: 0, jsonl_records: 1,508, nonempty_answer_records: 1,508, unique_run_label_query_id_sample_keys: 1,508
initial-logit-lenssmokingmirror/freeform/lens_yi/distinct_model_strings: 1, distinct_run_labels: 1, graded_n20_csv_data_rows: 20, graded_n3_csv_data_rows: 3, json_readouts: 20, nonempty_token_mass_records: 20
jacobian-directory-current-artifactssmokingmirror/freeform/lens_jac/baseline_condition_records: 1,040, distinct_model_strings: 47, distinct_record_run_labels: 52, files_per_run_label_directory: 20, json_readouts: 1,040, nonempty_baseline_token_mass_records: 1,020, nonempty_token_mass_records: 1,040, records_declaring_lens_jacobian_jvp: 1,020, records_without_lens_marker: 20, run_label_directories: 52
jacobian-grading-snapshotsmokingmirror/freeform/JACOBIAN-REMEASURE.md, smokingmirror/freeform/site/data/spectrum.jsondistinct_flagged_run_rows: 12, graded_question_records_all_rows: 840, graded_question_records_in_unflagged_rows: 753, instrument_suspect_run_rows: 6, provisional_run_rows: 1, rows_with_any_graded_question: 45, rows_with_at_least_five_graded_questions: 44, table_run_rows: 51, unflagged_run_rows: 39, ungraded_run_rows: 6, unparseable_question_records: 179
identity-instrument-versions—
descent-trajectoriessmokingmirror/freeform/descent_raw/depth_cap_terminals: 96, models: 2, other_terminal_statuses: 0, rung_json_records: 768, terminal_json_summaries: 96, total_json_files: 864, trajectories: 96
lineage-derived-evidencesmokingmirror/lineage-evidence/lineage-pairs-2026-08-30.csv, smokingmirror/lineage-evidence/lineage-code-envelopes-2026-08-30.csvcode_envelope_summary_rows: 14, model_pair_comparison_rows: 8, sum_joined_probes_in_code_envelope_rows: 24,161

Built 2026-10-08 14:45 UTC from the live research database and site catalogue by smokingmirror/reports/inventory/build_inventory.py (hop w-90faad5e). Raw record: inventory.json