FAILURE MAP

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 ↗
100840Executable case variants
20168Distinct failure mechanisms
302520Executed implementations
20168Open-access cases

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

Python · Standard library
REFERENCEFAILURE MECHANISMDOMAINACCESS
FA-98601

Prefix cache reuse crosses namespace boundaries · case 01

Matching token prefixes are reused across incompatible cache namespaces.

KV cache position mapping● Open access↗
FA-98602

Prefix cache reuse crosses namespace boundaries · case 02

Matching token prefixes are reused across incompatible cache namespaces.

KV cache position mapping◈ Members↗
FA-98603

Prefix cache reuse crosses namespace boundaries · case 03

Matching token prefixes are reused across incompatible cache namespaces.

KV cache position mapping◈ Members↗
FA-98604

Prefix cache reuse crosses namespace boundaries · case 04

Matching token prefixes are reused across incompatible cache namespaces.

KV cache position mapping◈ Members↗
FA-98605

Prefix cache reuse crosses namespace boundaries · case 05

Matching token prefixes are reused across incompatible cache namespaces.

KV cache position mapping◈ Members↗
FA-98606

Rewriting a logical position replaces its prior slot · case 01

A retry creates two slot entries for the same logical position.

KV cache position mapping● Open access↗
FA-98607

Rewriting a logical position replaces its prior slot · case 02

A retry creates two slot entries for the same logical position.

KV cache position mapping◈ Members↗
FA-98608

Rewriting a logical position replaces its prior slot · case 03

A retry creates two slot entries for the same logical position.

KV cache position mapping◈ Members↗
FA-98609

Rewriting a logical position replaces its prior slot · case 04

A retry creates two slot entries for the same logical position.

KV cache position mapping◈ Members↗
FA-98610

Rewriting a logical position replaces its prior slot · case 05

A retry creates two slot entries for the same logical position.

KV cache position mapping◈ Members↗
FA-98611

Slot allocation uses the lowest available physical slot · case 01

Allocation skips reusable holes and grows the cache unnecessarily.

KV cache position mapping● Open access↗
FA-98612

Slot allocation uses the lowest available physical slot · case 02

Allocation skips reusable holes and grows the cache unnecessarily.

KV cache position mapping◈ Members↗
FA-98613

Slot allocation uses the lowest available physical slot · case 03

Allocation skips reusable holes and grows the cache unnecessarily.

KV cache position mapping◈ Members↗
FA-98614

Slot allocation uses the lowest available physical slot · case 04

Allocation skips reusable holes and grows the cache unnecessarily.

KV cache position mapping◈ Members↗
FA-98615

Slot allocation uses the lowest available physical slot · case 05

Allocation skips reusable holes and grows the cache unnecessarily.

KV cache position mapping◈ Members↗
FA-98616

Batch row ownership follows request IDs · case 01

Cache rows are assigned by transient request list position.

KV cache position mapping● Open access↗
FA-98617

Batch row ownership follows request IDs · case 02

Cache rows are assigned by transient request list position.

KV cache position mapping◈ Members↗
FA-98618

Batch row ownership follows request IDs · case 03

Cache rows are assigned by transient request list position.

KV cache position mapping◈ Members↗
FA-98619

Batch row ownership follows request IDs · case 04

Cache rows are assigned by transient request list position.

KV cache position mapping◈ Members↗
FA-98620

Batch row ownership follows request IDs · case 05

Cache rows are assigned by transient request list position.

KV cache position mapping◈ Members↗
FA-98621

Cache trimming includes the exclusive end position · case 01

The first position beyond the requested half-open span remains in the cache.

KV cache position mapping● Open access↗
FA-98622

Cache trimming includes the exclusive end position · case 02

The first position beyond the requested half-open span remains in the cache.

KV cache position mapping◈ Members↗
FA-98623

Cache trimming includes the exclusive end position · case 03

The first position beyond the requested half-open span remains in the cache.

KV cache position mapping◈ Members↗
FA-98624

Cache trimming includes the exclusive end position · case 04

The first position beyond the requested half-open span remains in the cache.

KV cache position mapping◈ Members↗
FA-98625

Cache trimming includes the exclusive end position · case 05

The first position beyond the requested half-open span remains in the cache.

KV cache position mapping◈ Members↗
FA-98626

Multi-token append reserves capacity atomically · case 01

A request that does not fit partially consumes free cache slots.

KV cache position mapping● Open access↗
FA-98627

Multi-token append reserves capacity atomically · case 02

A request that does not fit partially consumes free cache slots.

KV cache position mapping◈ Members↗
FA-98628

Multi-token append reserves capacity atomically · case 03

A request that does not fit partially consumes free cache slots.

KV cache position mapping◈ Members↗
FA-98629

Multi-token append reserves capacity atomically · case 04

A request that does not fit partially consumes free cache slots.

KV cache position mapping◈ Members↗
FA-98630

Multi-token append reserves capacity atomically · case 05

A request that does not fit partially consumes free cache slots.

KV cache position mapping◈ Members↗
FA-98631

Reused slot handles reject stale generations · case 01

A handle from a prior owner resolves to a newly reused slot.

KV cache position mapping● Open access↗
FA-98632

Reused slot handles reject stale generations · case 02

A handle from a prior owner resolves to a newly reused slot.

KV cache position mapping◈ Members↗
FA-98633

Reused slot handles reject stale generations · case 03

A handle from a prior owner resolves to a newly reused slot.

KV cache position mapping◈ Members↗
FA-98634

Reused slot handles reject stale generations · case 04

A handle from a prior owner resolves to a newly reused slot.

KV cache position mapping◈ Members↗
FA-98635

Reused slot handles reject stale generations · case 05

A handle from a prior owner resolves to a newly reused slot.

KV cache position mapping◈ Members↗
FA-98636

Prefix reuse rounds down to complete cache blocks · case 01

A partial cache block is claimed as reusable.

KV cache position mapping● Open access↗
FA-98637

Prefix reuse rounds down to complete cache blocks · case 02

A partial cache block is claimed as reusable.

KV cache position mapping◈ Members↗
FA-98638

Prefix reuse rounds down to complete cache blocks · case 03

A partial cache block is claimed as reusable.

KV cache position mapping◈ Members↗
FA-98639

Prefix reuse rounds down to complete cache blocks · case 04

A partial cache block is claimed as reusable.

KV cache position mapping◈ Members↗
FA-98640

Prefix reuse rounds down to complete cache blocks · case 05

A partial cache block is claimed as reusable.

KV cache position mapping◈ Members↗
FA-98641

Quantization: Affine codes multiply by scale before rounding · case 01

Affine quantized codes are too large when scale exceeds one.

Model quantization arithmetic● Open access↗
FA-98642

Quantization: Affine codes multiply by scale before rounding · case 02

Affine quantized codes are too large when scale exceeds one.

Model quantization arithmetic◈ Members↗
FA-98643

Quantization: Affine codes multiply by scale before rounding · case 03

Affine quantized codes are too large when scale exceeds one.

Model quantization arithmetic◈ Members↗
FA-98644

Quantization: Affine codes multiply by scale before rounding · case 04

Affine quantized codes are too large when scale exceeds one.

Model quantization arithmetic◈ Members↗
FA-98645

Quantization: Affine codes multiply by scale before rounding · case 05

Affine quantized codes are too large when scale exceeds one.

Model quantization arithmetic◈ Members↗
FA-98646

Quantization: Zero point is added before dequantization scaling · case 01

The zero point is included in the value before dequantization scaling.

Model quantization arithmetic● Open access↗
FA-98647

Quantization: Zero point is added before dequantization scaling · case 02

The zero point is included in the value before dequantization scaling.

Model quantization arithmetic◈ Members↗
FA-98648

Quantization: Zero point is added before dequantization scaling · case 03

The zero point is included in the value before dequantization scaling.

Model quantization arithmetic◈ Members↗
FA-98649

Quantization: Zero point is added before dequantization scaling · case 04

The zero point is included in the value before dequantization scaling.

Model quantization arithmetic◈ Members↗
FA-98650

Quantization: Zero point is added before dequantization scaling · case 05

The zero point is included in the value before dequantization scaling.

Model quantization arithmetic◈ Members↗
FA-98651

Quantization: Per-channel scales follow the wrong axis · case 01

Per-channel scales follow the wrong axis.

Model quantization arithmetic● Open access↗
FA-98652

Quantization: Per-channel scales follow the wrong axis · case 02

Per-channel scales follow the wrong axis.

Model quantization arithmetic◈ Members↗
FA-98653

Quantization: Per-channel scales follow the wrong axis · case 03

Per-channel scales follow the wrong axis.

Model quantization arithmetic◈ Members↗
FA-98654

Quantization: Per-channel scales follow the wrong axis · case 04

Per-channel scales follow the wrong axis.

Model quantization arithmetic◈ Members↗
FA-98655

Quantization: Per-channel scales follow the wrong axis · case 05

Per-channel scales follow the wrong axis.

Model quantization arithmetic◈ Members↗
FA-98656

Quantization: Per-channel zero points broadcast by row · case 01

Per-channel zero points broadcast by row.

Model quantization arithmetic● Open access↗
FA-98657

Quantization: Per-channel zero points broadcast by row · case 02

Per-channel zero points broadcast by row.

Model quantization arithmetic◈ Members↗
FA-98658

Quantization: Per-channel zero points broadcast by row · case 03

Per-channel zero points broadcast by row.

Model quantization arithmetic◈ Members↗
FA-98659

Quantization: Per-channel zero points broadcast by row · case 04

Per-channel zero points broadcast by row.

Model quantization arithmetic◈ Members↗
FA-98660

Quantization: Per-channel zero points broadcast by row · case 05

Per-channel zero points broadcast by row.

Model quantization arithmetic◈ Members↗
FA-98661

Quantization: Midpoint rounding uses half-up · case 01

Midpoint rounding uses half-up.

Model quantization arithmetic● Open access↗
FA-98662

Quantization: Midpoint rounding uses half-up · case 02

Midpoint rounding uses half-up.

Model quantization arithmetic◈ Members↗
FA-98663

Quantization: Midpoint rounding uses half-up · case 03

Midpoint rounding uses half-up.

Model quantization arithmetic◈ Members↗
FA-98664

Quantization: Midpoint rounding uses half-up · case 04

Midpoint rounding uses half-up.

Model quantization arithmetic◈ Members↗
FA-98665

Quantization: Midpoint rounding uses half-up · case 05

Midpoint rounding uses half-up.

Model quantization arithmetic◈ Members↗
FA-98666

Quantization: Negative midpoint rounding is asymmetric · case 01

Negative midpoint rounding is asymmetric.

Model quantization arithmetic● Open access↗
FA-98667

Quantization: Negative midpoint rounding is asymmetric · case 02

Negative midpoint rounding is asymmetric.

Model quantization arithmetic◈ Members↗
FA-98668

Quantization: Negative midpoint rounding is asymmetric · case 03

Negative midpoint rounding is asymmetric.

Model quantization arithmetic◈ Members↗
FA-98669

Quantization: Negative midpoint rounding is asymmetric · case 04

Negative midpoint rounding is asymmetric.

Model quantization arithmetic◈ Members↗
FA-98670

Quantization: Negative midpoint rounding is asymmetric · case 05

Negative midpoint rounding is asymmetric.

Model quantization arithmetic◈ Members↗
FA-98671

Quantization: Clipping occurs before zero-point shift · case 01

Clipping occurs before zero-point shift.

Model quantization arithmetic● Open access↗
FA-98672

Quantization: Clipping occurs before zero-point shift · case 02

Clipping occurs before zero-point shift.

Model quantization arithmetic◈ Members↗
FA-98673

Quantization: Clipping occurs before zero-point shift · case 03

Clipping occurs before zero-point shift.

Model quantization arithmetic◈ Members↗
FA-98674

Quantization: Clipping occurs before zero-point shift · case 04

Clipping occurs before zero-point shift.

Model quantization arithmetic◈ Members↗
FA-98675

Quantization: Clipping occurs before zero-point shift · case 05

Clipping occurs before zero-point shift.

Model quantization arithmetic◈ Members↗
FA-98676

Quantization: Quantized upper code is excluded · case 01

Quantized upper code is excluded.

Model quantization arithmetic● Open access↗
FA-98677

Quantization: Quantized upper code is excluded · case 02

Quantized upper code is excluded.

Model quantization arithmetic◈ Members↗
FA-98678

Quantization: Quantized upper code is excluded · case 03

Quantized upper code is excluded.

Model quantization arithmetic◈ Members↗
FA-98679

Quantization: Quantized upper code is excluded · case 04

Quantized upper code is excluded.

Model quantization arithmetic◈ Members↗
FA-98680

Quantization: Quantized upper code is excluded · case 05

Quantized upper code is excluded.

Model quantization arithmetic◈ Members↗
FA-98681

Quantization: Fixed-point rescale truncates instead of rounding · case 01

Fixed-point rescale truncates instead of rounding.

Model quantization arithmetic● Open access↗
FA-98682

Quantization: Fixed-point rescale truncates instead of rounding · case 02

Fixed-point rescale truncates instead of rounding.

Model quantization arithmetic◈ Members↗
FA-98683

Quantization: Fixed-point rescale truncates instead of rounding · case 03

Fixed-point rescale truncates instead of rounding.

Model quantization arithmetic◈ Members↗
FA-98684

Quantization: Fixed-point rescale truncates instead of rounding · case 04

Fixed-point rescale truncates instead of rounding.

Model quantization arithmetic◈ Members↗
FA-98685

Quantization: Fixed-point rescale truncates instead of rounding · case 05

Fixed-point rescale truncates instead of rounding.

Model quantization arithmetic◈ Members↗
FA-98686

Quantization: Left rescale shifts in the wrong direction · case 01

Left rescale shifts in the wrong direction.

Model quantization arithmetic● Open access↗
FA-98687

Quantization: Left rescale shifts in the wrong direction · case 02

Left rescale shifts in the wrong direction.

Model quantization arithmetic◈ Members↗
FA-98688

Quantization: Left rescale shifts in the wrong direction · case 03

Left rescale shifts in the wrong direction.

Model quantization arithmetic◈ Members↗
FA-98689

Quantization: Left rescale shifts in the wrong direction · case 04

Left rescale shifts in the wrong direction.

Model quantization arithmetic◈ Members↗
FA-98690

Quantization: Left rescale shifts in the wrong direction · case 05

Left rescale shifts in the wrong direction.

Model quantization arithmetic◈ Members↗
FA-98691

Quantization: Multiplier rescale rounds before multiplying · case 01

Multiplier rescale rounds before multiplying.

Model quantization arithmetic● Open access↗
FA-98692

Quantization: Multiplier rescale rounds before multiplying · case 02

Multiplier rescale rounds before multiplying.

Model quantization arithmetic◈ Members↗
FA-98693

Quantization: Multiplier rescale rounds before multiplying · case 03

Multiplier rescale rounds before multiplying.

Model quantization arithmetic◈ Members↗
FA-98694

Quantization: Multiplier rescale rounds before multiplying · case 04

Multiplier rescale rounds before multiplying.

Model quantization arithmetic◈ Members↗
FA-98695

Quantization: Multiplier rescale rounds before multiplying · case 05

Multiplier rescale rounds before multiplying.

Model quantization arithmetic◈ Members↗
FA-98696

Quantization: Input zero point correction skips right-shift rescaling · case 01

Centered accumulator values are left at the wrong scale.

Model quantization arithmetic● Open access↗
FA-98697

Quantization: Input zero point correction skips right-shift rescaling · case 02

Centered accumulator values are left at the wrong scale.

Model quantization arithmetic◈ Members↗
FA-98698

Quantization: Input zero point correction skips right-shift rescaling · case 03

Centered accumulator values are left at the wrong scale.

Model quantization arithmetic◈ Members↗
FA-98699

Quantization: Input zero point correction skips right-shift rescaling · case 04

Centered accumulator values are left at the wrong scale.

Model quantization arithmetic◈ Members↗
FA-98700

Quantization: Input zero point correction skips right-shift rescaling · case 05

Centered accumulator values are left at the wrong scale.

Model quantization arithmetic◈ Members↗

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 ↗