Understand the failure.
Verify the repair.
Small, reproducible software failures. The broken implementation, the fix that didn’t work, and the one that passed—preserved together.
Explore the cases ↓How results are verified ↗WHAT THE ARCHIVE CONTAINS
100840 executable cases. 20168 are open.
Every case records the implementation that fails, the fix that did not work, and the repair that passed its checks—with recorded outputs and source hashes. This release adds 100840 cases across 20168 failure mechanisms and 254 domains.
The open tier gives you the failure and the unsuccessful fix for one case in every mechanism. The remaining 80672 cases, 5 variants per mechanism, are member-only: the verified repair, its recorded checks, and the full fixture suite are held in the member archive. Read the methodology ↗
A RECORD OF WHAT WENT WRONG
Browse the archive / 100840
MinHash Jaccard estimation: offset added after reduction · case 01
Signature values exceed the field and disagree with producers that reduce the full expression.
MinHash Jaccard estimation: offset added after reduction · case 02
Signature values exceed the field and disagree with producers that reduce the full expression.
MinHash Jaccard estimation: offset added after reduction · case 03
Signature values exceed the field and disagree with producers that reduce the full expression.
MinHash Jaccard estimation: offset added after reduction · case 04
Signature values exceed the field and disagree with producers that reduce the full expression.
MinHash Jaccard estimation: offset added after reduction · case 05
Signature values exceed the field and disagree with producers that reduce the full expression.
MinHash Jaccard estimation: one empty set reports undefined similarity · case 01
A document compared with an empty one is treated as undefined instead of disjoint.
MinHash Jaccard estimation: one empty set reports undefined similarity · case 02
A document compared with an empty one is treated as undefined instead of disjoint.
MinHash Jaccard estimation: one empty set reports undefined similarity · case 03
A document compared with an empty one is treated as undefined instead of disjoint.
MinHash Jaccard estimation: one empty set reports undefined similarity · case 04
A document compared with an empty one is treated as undefined instead of disjoint.
MinHash Jaccard estimation: one empty set reports undefined similarity · case 05
A document compared with an empty one is treated as undefined instead of disjoint.
MinHash LSH banding: band key omits the band index · case 01
Two signatures that agree on different bands are paired although no band matches.
MinHash LSH banding: band key omits the band index · case 02
Two signatures that agree on different bands are paired although no band matches.
MinHash LSH banding: band key omits the band index · case 03
Two signatures that agree on different bands are paired although no band matches.
MinHash LSH banding: band key omits the band index · case 04
Two signatures that agree on different bands are paired although no band matches.
MinHash LSH banding: band key omits the band index · case 05
Two signatures that agree on different bands are paired although no band matches.
MinHash LSH banding: band slice uses the band count as width · case 01
Band boundaries shift whenever b differs from r, so candidate pairs change.
MinHash LSH banding: band slice uses the band count as width · case 02
Band boundaries shift whenever b differs from r, so candidate pairs change.
MinHash LSH banding: band slice uses the band count as width · case 03
Band boundaries shift whenever b differs from r, so candidate pairs change.
MinHash LSH banding: band slice uses the band count as width · case 04
Band boundaries shift whenever b differs from r, so candidate pairs change.
MinHash LSH banding: band slice uses the band count as width · case 05
Band boundaries shift whenever b differs from r, so candidate pairs change.
MinHash LSH banding: overlong signatures accepted · case 01
Signatures produced with a different hash count are compared on a prefix instead of being rejected.
MinHash LSH banding: overlong signatures accepted · case 02
Signatures produced with a different hash count are compared on a prefix instead of being rejected.
MinHash LSH banding: overlong signatures accepted · case 03
Signatures produced with a different hash count are compared on a prefix instead of being rejected.
MinHash LSH banding: overlong signatures accepted · case 04
Signatures produced with a different hash count are compared on a prefix instead of being rejected.
MinHash LSH banding: overlong signatures accepted · case 05
Signatures produced with a different hash count are compared on a prefix instead of being rejected.
MinHash LSH banding: threshold swaps bands and rows · case 01
Operators tune b and r against a threshold that describes a different S-curve.
MinHash LSH banding: threshold swaps bands and rows · case 02
Operators tune b and r against a threshold that describes a different S-curve.
MinHash LSH banding: threshold swaps bands and rows · case 03
Operators tune b and r against a threshold that describes a different S-curve.
MinHash LSH banding: threshold swaps bands and rows · case 04
Operators tune b and r against a threshold that describes a different S-curve.
MinHash LSH banding: threshold swaps bands and rows · case 05
Operators tune b and r against a threshold that describes a different S-curve.
MinHash LSH banding: pairs reported in both orders · case 01
The verifier compares every candidate twice, once per ordering.
MinHash LSH banding: pairs reported in both orders · case 02
The verifier compares every candidate twice, once per ordering.
MinHash LSH banding: pairs reported in both orders · case 03
The verifier compares every candidate twice, once per ordering.
MinHash LSH banding: pairs reported in both orders · case 04
The verifier compares every candidate twice, once per ordering.
MinHash LSH banding: pairs reported in both orders · case 05
The verifier compares every candidate twice, once per ordering.
K-minimum-values distinct counting: repeated hashes occupy sketch slots · case 01
A stream that repeats one small hash looks like it has few distinct values beyond it, inflating the estimate.
K-minimum-values distinct counting: repeated hashes occupy sketch slots · case 02
A stream that repeats one small hash looks like it has few distinct values beyond it, inflating the estimate.
K-minimum-values distinct counting: repeated hashes occupy sketch slots · case 03
A stream that repeats one small hash looks like it has few distinct values beyond it, inflating the estimate.
K-minimum-values distinct counting: repeated hashes occupy sketch slots · case 04
A stream that repeats one small hash looks like it has few distinct values beyond it, inflating the estimate.
K-minimum-values distinct counting: repeated hashes occupy sketch slots · case 05
A stream that repeats one small hash looks like it has few distinct values beyond it, inflating the estimate.
K-minimum-values distinct counting: estimator uses k instead of k-1 · case 01
Distinct counts are biased upward by a factor k/(k-1).
K-minimum-values distinct counting: estimator uses k instead of k-1 · case 02
Distinct counts are biased upward by a factor k/(k-1).
K-minimum-values distinct counting: estimator uses k instead of k-1 · case 03
Distinct counts are biased upward by a factor k/(k-1).
K-minimum-values distinct counting: estimator uses k instead of k-1 · case 04
Distinct counts are biased upward by a factor k/(k-1).
K-minimum-values distinct counting: estimator uses k instead of k-1 · case 05
Distinct counts are biased upward by a factor k/(k-1).
K-minimum-values distinct counting: full sketch reported as an exact count · case 01
A sketch holding exactly k values reports k, although the true distinct count may be far larger.
K-minimum-values distinct counting: full sketch reported as an exact count · case 02
A sketch holding exactly k values reports k, although the true distinct count may be far larger.
K-minimum-values distinct counting: full sketch reported as an exact count · case 03
A sketch holding exactly k values reports k, although the true distinct count may be far larger.
K-minimum-values distinct counting: full sketch reported as an exact count · case 04
A sketch holding exactly k values reports k, although the true distinct count may be far larger.
K-minimum-values distinct counting: full sketch reported as an exact count · case 05
A sketch holding exactly k values reports k, although the true distinct count may be far larger.
K-minimum-values distinct counting: union keeps shared hashes twice · case 01
Values present in both sketches occupy two union slots, overstating the union and its overlap.
K-minimum-values distinct counting: union keeps shared hashes twice · case 02
Values present in both sketches occupy two union slots, overstating the union and its overlap.
K-minimum-values distinct counting: union keeps shared hashes twice · case 03
Values present in both sketches occupy two union slots, overstating the union and its overlap.
K-minimum-values distinct counting: union keeps shared hashes twice · case 04
Values present in both sketches occupy two union slots, overstating the union and its overlap.
K-minimum-values distinct counting: union keeps shared hashes twice · case 05
Values present in both sketches occupy two union slots, overstating the union and its overlap.
K-minimum-values distinct counting: overlap counted beyond the union sketch · case 01
Shared values larger than the union's k-th value inflate the intersection estimate.
K-minimum-values distinct counting: overlap counted beyond the union sketch · case 02
Shared values larger than the union's k-th value inflate the intersection estimate.
K-minimum-values distinct counting: overlap counted beyond the union sketch · case 03
Shared values larger than the union's k-th value inflate the intersection estimate.
K-minimum-values distinct counting: overlap counted beyond the union sketch · case 04
Shared values larger than the union's k-th value inflate the intersection estimate.
K-minimum-values distinct counting: overlap counted beyond the union sketch · case 05
Shared values larger than the union's k-th value inflate the intersection estimate.
K-minimum-values distinct counting: intersection truncated · case 01
Intersection estimates are biased down by up to one.
K-minimum-values distinct counting: intersection truncated · case 02
Intersection estimates are biased down by up to one.
K-minimum-values distinct counting: intersection truncated · case 03
Intersection estimates are biased down by up to one.
K-minimum-values distinct counting: intersection truncated · case 04
Intersection estimates are biased down by up to one.
K-minimum-values distinct counting: intersection truncated · case 05
Intersection estimates are biased down by up to one.
Cuckoo filter with partial-key relocation: zero fingerprint stored · case 01
An item whose fingerprint byte is zero is stored as the empty-slot marker and vanishes in packed tables.
Cuckoo filter with partial-key relocation: zero fingerprint stored · case 02
An item whose fingerprint byte is zero is stored as the empty-slot marker and vanishes in packed tables.
Cuckoo filter with partial-key relocation: zero fingerprint stored · case 03
An item whose fingerprint byte is zero is stored as the empty-slot marker and vanishes in packed tables.
Cuckoo filter with partial-key relocation: zero fingerprint stored · case 04
An item whose fingerprint byte is zero is stored as the empty-slot marker and vanishes in packed tables.
Cuckoo filter with partial-key relocation: zero fingerprint stored · case 05
An item whose fingerprint byte is zero is stored as the empty-slot marker and vanishes in packed tables.
Cuckoo filter with partial-key relocation: alternate bucket hashed from the full item · case 01
Relocated fingerprints move to buckets that lookups never check, producing false negatives.
Cuckoo filter with partial-key relocation: alternate bucket hashed from the full item · case 02
Relocated fingerprints move to buckets that lookups never check, producing false negatives.
Cuckoo filter with partial-key relocation: alternate bucket hashed from the full item · case 03
Relocated fingerprints move to buckets that lookups never check, producing false negatives.
Cuckoo filter with partial-key relocation: alternate bucket hashed from the full item · case 04
Relocated fingerprints move to buckets that lookups never check, producing false negatives.
Cuckoo filter with partial-key relocation: alternate bucket hashed from the full item · case 05
Relocated fingerprints move to buckets that lookups never check, producing false negatives.
Cuckoo filter with partial-key relocation: one relocation fewer than configured · case 01
Inserts report full although the configured kick budget would have placed them.
Cuckoo filter with partial-key relocation: one relocation fewer than configured · case 02
Inserts report full although the configured kick budget would have placed them.
Cuckoo filter with partial-key relocation: one relocation fewer than configured · case 03
Inserts report full although the configured kick budget would have placed them.
Cuckoo filter with partial-key relocation: one relocation fewer than configured · case 04
Inserts report full although the configured kick budget would have placed them.
Cuckoo filter with partial-key relocation: one relocation fewer than configured · case 05
Inserts report full although the configured kick budget would have placed them.
Cuckoo filter with partial-key relocation: failed insert keeps partial relocations · case 01
A failed insert silently evicts some other item's fingerprint, creating a false negative.
Cuckoo filter with partial-key relocation: failed insert keeps partial relocations · case 02
A failed insert silently evicts some other item's fingerprint, creating a false negative.
Cuckoo filter with partial-key relocation: failed insert keeps partial relocations · case 03
A failed insert silently evicts some other item's fingerprint, creating a false negative.
Cuckoo filter with partial-key relocation: failed insert keeps partial relocations · case 04
A failed insert silently evicts some other item's fingerprint, creating a false negative.
Cuckoo filter with partial-key relocation: failed insert keeps partial relocations · case 05
A failed insert silently evicts some other item's fingerprint, creating a false negative.
Cuckoo filter with partial-key relocation: delete removes every copy · case 01
Deleting one of several identical inserts removes all of them.
Cuckoo filter with partial-key relocation: delete removes every copy · case 02
Deleting one of several identical inserts removes all of them.
Cuckoo filter with partial-key relocation: delete removes every copy · case 03
Deleting one of several identical inserts removes all of them.
Cuckoo filter with partial-key relocation: delete removes every copy · case 04
Deleting one of several identical inserts removes all of them.
Cuckoo filter with partial-key relocation: delete removes every copy · case 05
Deleting one of several identical inserts removes all of them.
Cuckoo filter with partial-key relocation: lookup checks only the primary bucket · case 01
Items placed in their alternate bucket are reported absent.
Cuckoo filter with partial-key relocation: lookup checks only the primary bucket · case 02
Items placed in their alternate bucket are reported absent.
Cuckoo filter with partial-key relocation: lookup checks only the primary bucket · case 03
Items placed in their alternate bucket are reported absent.
Cuckoo filter with partial-key relocation: lookup checks only the primary bucket · case 04
Items placed in their alternate bucket are reported absent.
Cuckoo filter with partial-key relocation: lookup checks only the primary bucket · case 05
Items placed in their alternate bucket are reported absent.
DGIM sliding-window bit counting: merged bucket keeps the older timestamp · case 01
Merged buckets expire too early and the window count drops below the true value.
DGIM sliding-window bit counting: merged bucket keeps the older timestamp · case 02
Merged buckets expire too early and the window count drops below the true value.
DGIM sliding-window bit counting: merged bucket keeps the older timestamp · case 03
Merged buckets expire too early and the window count drops below the true value.
DGIM sliding-window bit counting: merged bucket keeps the older timestamp · case 04
Merged buckets expire too early and the window count drops below the true value.
DGIM sliding-window bit counting: merged bucket keeps the older timestamp · case 05
Merged buckets expire too early and the window count drops below the true value.
INSPECTABLE BY DESIGN
Every result has a runnable source.
Runnable implementations with recorded outputs, source hashes, and explicit contracts. Related variants share a failure mechanism and belong together in evaluation splits.
Read the methodology ↗