EXP-01 / Provider record
7J6JV...PQ1X4
Reproducibility
Reproducibility
Unrated
No rerun has ever been performed here.
Attested traces carry zero reproducibility weight: TEE attestation proves a binary ran, not that its output reproduces. They are reported under traces_untested.
Estimator, published in full
A score whose method is withheld is a claim, not a measurement. The constants below are the ones this figure was computed with, so it can be recomputed and disagreed with.
- Estimator
- wilson_score_lower_bound
- Confidence z
- 1.96two-sided 95%
- Recency half-life
- 30days -- an old comparison counts for less
- Min effective samples
- 5below this, no rate is stated at all
- Tier weights
- Proven1.00
- Incident rerun1.00
- Sampled0.60
- Attested0.00
Totals
Traces committed
15
Re-run and compared
0
Never re-run
15
2 carry attestation only
Coverage
0.0%
share of committed traces with a comparison
Matches
0
Mismatches
0
Incidents
0
0 open / 0 upheld / 0 rejected
Pending challenge
0
traces inside an open challenge window
Distribution by tier
The tiers do not carry equal weight and are not averaged as if they did. A tier that was never re-run is marked excluded rather than scored, because a tier with no comparisons has no rate -- not a rate of zero.
| Tier | Reproducibility | Comparisons | Matches | Mismatches | Traces | Never re-run | Effective |
|---|---|---|---|---|---|---|---|
| Attested | Excluded attestation is not a rerun; these traces carry no reproducibility evidence unless challenged | 0 | 0 | 0 | 2 | 2 | 0.0 |
| Sampled | Unrated No rerun has ever been performed here. | 0 | 0 | 0 | 13 | 13 | 0.0 |
Distribution by model plate
The same provider, split by the model each trace was committed against. 0 of 2 plates carry enough evidence to be rated; the rest are listed unrated so that the aggregate is not read as covering them.
| Model plate | Reproducibility | Comparisons | Matches | Mismatches | Traces | Never re-run | Effective |
|---|---|---|---|---|---|---|---|
nucleate/call-v1 6QF6r6...nu8SfJ | Unrated No rerun has ever been performed here. | 0 | 0 | 0 | 2 | 2 | 0.0 |
condense/qwen2.5-1.5b-instruct-q4_k_m DrdGNS...SMj7fD | Unrated No rerun has ever been performed here. | 0 | 0 | 0 | 13 | 13 | 0.0 |
Attribution
An average hides local failure. One rate over every comparison cannot say which model produced the mismatches or how much of the failure weight sits on a single plate, so the total, the distribution and the attribution are reported as three separate answers instead of one.
Evidence by kind
No output comparison has been recorded for this provider. That is not a score of zero -- nothing has been re-run, so there is nothing to attribute.
Where it landed
- Best plate
- No plate carries enough evidence to be ranked.
- Worst plate
- No plate carries enough evidence to be ranked.
- Top failure plateshare of weighted failure
- No failure has been attributed to any plate.
- Latest mismatch
- No comparison has come back mismatched.
- Plates rated
- 0 of 2
Concentration
Evidence herfindahl
--
no re-run has landed anywhere yet
Failure herfindahl
--
no failure has been recorded
Plates with evidence
0
Plates without evidence
2
registered, never re-run
evidence_herfindahl near 1.0 means nearly all reruns landed on a single model, so the aggregate score does not describe the others
Reproducibility history
One row is one output comparison. The weight column shows how the row entered the score: tier weight multiplied by recency weight, both printed, so no row contributes an amount the reader cannot see.
Nothing recorded
No output comparison has been performed for this provider.
Committed traces that were never re-run are counted under totals.traces_untested, not here.
Every row is one output comparison that actually happened. Traces that were never re-run do not appear here and are counted under totals.traces_untested.
Registered model plates
| Model | Version | Status | Bond | Traces | Incidents | Mismatches | Registered |
|---|---|---|---|---|---|---|---|
nucleate/call-v1 6QF6r6...nu8SfJ | 2026-10-05 | Active | 1,000,000KORTX | 2 | 0 | 0 | 2026-10-06 05:55 UTC |
condense/qwen2.5-1.5b-instruct-q4_k_m DrdGNS...SMj7fD | 2026-10-05 | Active | 1,000,000KORTX | 13 | 0 | 0 | 2026-10-06 05:55 UTC |
Indexer ok / read through slot 453,947,649 / 2026-10-06 15:54 UTC