$19.499Observed · provisional
0.682Public leaderboard
251of 5,463 teams
0.765Same ranked snapshot
-0.083Balanced accuracy
18:00Daily · America/New_York
01 / Capital ledger
USD · FAILED ATTEMPTS INCLUDED ONCE| Campaign / item | Observed $ | As of | Billing state | Source |
|---|---|---|---|---|
| Qwen one-pass model | 4.856 | 11 Nov 08:58 EST | Completed; posted charges by exact instance | Receipt ↗ |
| Qwen failed hosts | 0.899 | 11 Nov 08:58 EST | Two failed hosts; included once by exact instance | Receipt ↗ |
| Qwen three-pass run | 13.206 | 11 Nov 08:58 EST | Completed; instance absent in saved shutdown receipt; posted charges may settle | Receipt ↗ |
| GLM CUDA pilot | 0.538 | 10 Nov 14:25 EST | Four attempts; rentals destroyed; provisional | Receipt ↗ |
QWEN LIMIT$18.961 / $24
GLM LIMIT$0.538 / $10
Saved Qwen posted charges: 11 Nov 08:58 EST; GLM cached 10 Nov 14:25 EST. Provider settlement and later GLM charges unknown. Limits are separate; not available account balances.
02 / Scored result tape
PUBLIC LEADERBOARD / NOT VALIDATION| Entry | Score | Date | Source |
|---|---|---|---|
| Previous public ensemble | 0.651 | 27 Oct | Receipt ↗ |
| Public reproduction | 0.681 | 10 Nov | Receipt ↗ |
| Matched public-only control | 0.682 | 10 Nov | Receipt ↗ |
Matched additions / vs 0.682 public control
| Member | Score | Δ |
|---|---|---|
| +10% old Qwen | 0.678 | -0.004 |
| +10% tokenizer | 0.675 | -0.007 |
| + token and byte | 0.669 | -0.013 |
| 0.683 recipe control | 0.679 | -0.003 |
| +10% full-pool Qwen | 0.675 | -0.004 |
Scored progression / zero-to-one scale
+0.031
Saved 11 Nov 03:00 EST receipts: 0.683-title control scored 0.679; full-pool Qwen 0.675, down 0.004 versus matched control. Best unchanged 0.682. Three-pass and newer CPU proxies are not board scores.
03 / Experiment blotter
NAMED BASELINES · SOURCE RECEIPTS · MEASURED OUTCOMES| # | Experiment | Evidence type | Result | Baseline → score | Interpretation / limits | Source |
|---|---|---|---|---|---|---|
| 01 | Full-pool Qwen | validation | +3.37 pp | 0.587169 → 0.620882 | Standalone: five prompt folds, one seed. Trainer and exposure changed together. | Receipt ↗ |
| 02 | Mistral diversity lead | validation | +1.78 pp | 0.622322 → 0.640127 | Older full-stack reference, cross-fitted blend weight; author-supplied predictions, not fully nested. Own full training staged. | Receipt ↗ |
| 03 | OLMo diversity lead | validation | +0.89 pp | 0.622322 → 0.631228 | Same older reference; author-supplied predictions and cross-fitted weight. Interval crosses zero; own weights not reproduced. | Receipt ↗ |
| 04 | Liquid frozen encoder | validation | −0.74 pp | 0.543060 → 0.535650 | 5% blend on old public reference; exact checkpoint and all 2,699 rows. Full-backbone training is a separate experiment. | Receipt ↗ |
| 05 | Pruned GLM | engineering | CUDA works | — | 45 layers, 62 exact shard hashes. Three passages at 0.743–0.794 seconds; 36.80 GiB peak reserved. First eight experts, not calibrated. No classifier accuracy yet. | Receipt ↗ |
| 06 | Qwen three-pass | validation | −1.19 pp | 0.620882 → 0.609021 | Completed five folds and final model, saved 11 Nov 08:52 EST. Standalone declined versus one-pass; 10% replacement gained 0.003341 on learned-group proxy. Not a leaderboard result; interval spans zero. | Receipt ↗ |
| 07 | Liquid supervised encoder | validation | +0.44 pp | 0.543060 → 0.547502 | Three seeds, five prompt folds; 5% blend on older public reference gained 0.004442. Explored share, not a board score. No active-job claim. | Receipt ↗ |
| 08 | Complementary CPU corrections | validation | +0.15 pp | 0.757619 → 0.759100 | Saved 11 Nov proxy: combined Qwen/GPT and byte corrections, complete layout; interval spans zero. CPU replay ready; exact dual-T4 and board pending. Paired-source style also replayed, no board score. | Receipt ↗ |
No experiments match. Clear search or select ALL.
04 / Next bets
RESEARCH QUEUE / NOT NEW SPEND AUTHORIZATION- 01Verify exact full-inference public control and three-pass Qwen comparison; title scores are not ours.
- 02Confirm CPU correction transfer and exact dual-T4 replay before score claims.
- 03Prove GLM on dual T4, then calibrate experts on competition prose.
- 04Compare completed Liquid seeds and creative specialists against current matched ensemble.
05 / Accounting & evidence notes
READ THE TIMESTAMPSSelected current rental campaigns and research results — not an all-time account reconciliation.
Separate historical API usage: $6.9033 + $0.1963 reserve. Receipt ↗
Basis, exclusions & comparison limits
Separate historical API project; not included in the rental subtotal. Earlier pilot, local electricity, subscriptions, free Kaggle quota and other older rentals are not reconciled here.
- Percentage-point validation gains use different named baselines. Never add them or compare them with leaderboard gains.
- Validation shares explored on shared prompts; public-model training exposure not independently established.
- No all-time spend total is claimed. Historical API usage is separate; other older work is not reconciled.
- Project budget caps are separate limits, not money spent or a fungible account balance.
- Creative collection saved receipts as of 2026-11-10T23:00:18.157862+00:00: returned usage $6.051409, unknown-charge reserve $0.175014, separate $10 cap. Includes failed attempts; not in rental subtotal and not a settled invoice. See reports/scorecard/creative-usage.json.
- New saved receipts reviewed 11 Nov: Qwen campaign $18.961 posted by instance, replacing (not adding to) prior $8.047 campaign snapshot; separate GLM $0.538 cached 10 Nov. Selected rental subtotal $19.499; Qwen cap $24, GLM cap $10. Not all-time spend or settled invoice.
- Creative wave two summary adds $2.727149 usage-or-reserve to wave one $6.226423, combined $8.953572 under $10 cap. New itemized actual/reserve split was not reconciled; preserve wave-one actual $6.051409 and reserve $0.175014 as cached 10 Nov, not a new total.
- Padding-dependent creative results superseded: corrected joint specialists gained 0.001111 on older reference; validation only, not a leaderboard gain.
- Source gap: fresh provider billing and remote process state not queried. Local process list checked at review; no relevant local research process matched. Saved shutdown evidence is dated, not a fresh provider query. No Kaggle or paid calls.
- Current status and submission ledger lag newer original receipts: best remains 0.682, board rank and completed three-pass model use the dated receipts linked here.