Latent Defect Detection Patents: Who Leads, Where the Gaps Are 2026
- One filer holds 25 of 66 records and the top five together account for 78.8% of all records in scope — this field is concentrated, not fragmented.
- Filings peaked at 15 in 2022 then fell to 1 by 2024, a -91% drop across that three-year span, though 2025-26 counts are still filling in as publications lag filing.
- G01R electric measurement covers 53.0% of records while AI-based computing (G06N) already touches 12.1% — the statistical core is crowded, the ML overlay is not.
Filing growth compares 2021 (11 records) with 2024 (1) — a three-year span. 2024 is the most recent year we treat as complete: publication lags filing by roughly 18 months, so 2025 onwards are still filling in and any growth rate that ends there would understate the field. Top-5 share is the combined record count of the five largest assignees divided by all 66 records in scope (CR5), not by the ranked leaders only.
What this patent set covers
This landscape tracks patent activity at the intersection of latent defect screening and statistical outlier detection in semiconductor test and reliability engineering — part average testing, spatial wafer-map signatures, and rules for flagging die that pass functional test but carry a latent reliability risk. The search combines defect-screening terms with specific technical routes: parametric outlier scoring, spatial signatures, wafer map pattern recognition, burn-in alternatives, zero-defect targets and screening rule design.
Sixty-six published records sit in scope between 2015 and the 2026-07-31 cut-off. That is a small, specialised corpus rather than a broad technology field, which makes the concentration at the top of the assignee ranking more informative than it would be in a larger dataset.
Filing trend and technology composition
Two views of the same 66 records: how filing activity moved year over year, and which IPC subclasses the claims actually sit in.
Filing trend: a sharp rise and fall
Filings were at zero in 2017, climbed to a peak of 15 in 2022, then dropped to 1 by 2024 — a -91% change across that span. Treat 2025 and 2026 counts as incomplete rather than a continued decline, since publication typically lags filing by around 18 months.
Where the claims sit
G01R (electric and magnetic measurement) appears in 53.0% of the 66 records, and H01L (semiconductor devices) in 25.8% — the traditional test-and-measure core. G06N (AI-based computing) already reaches 12.1% and G06T (image processing) 9.1%, showing the field's statistical methods are increasingly framed as machine-learning claims, not just as classical parametric rules.
Shares are the percentage of the 66 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Go deeper on Latent Defect and Outlier Detection with Eureka
This page is one run against one query. Ask Eureka your own question about latent defect and outlier detection and every answer comes back with the patent numbers behind it.
Try EurekaRepresentative filing
System and method for z-pat defect-guided statistical outlier detection of semiconductor reliability failures
The filing describes receiving electrical test bin data across a wafer lot, running Z-direction Part Average Testing to flag statistical outliers, and correlating that output against fab characterization data generated during wafer fabrication — tying electrical outlier flags back to physical process signals rather than treating test data in isolation.Filed by KLA Corporation, published 2022-12-08.
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US6789032B2 | Method of statistical binning for reliability selection | 63 |
| 2 | WO2012037456A1 | Functional genomics assay for characterizing pluripotent stem cell utility and safety | 40 |
| 3 | US20130296183A1 | Functional genomics assay for characterizing pluripotent stem cell utility and safety | 34 |
| 4 | US20150148040A1 | Anomaly correlation mechanism for analysis of handovers in a communication network | 32 |
| 5 | US20210215753A1 | Advanced in-line part average testing | 25 |
| 6 | US20030120445A1 | Method of statistical binning for reliability selection | 25 |
| 7 | US7062415B2 | Parametric outlier detection | 22 |
| 8 | WO2007098426A2 | Methods and apparatus for data analysis | 14 |
| 9 | US20160364745A1 | Outlier data detection | 13 |
| 10 | WO2020206464A1 | Assay accuracy and reliability improvement | 10 |
Citation counts reflect prominence within this searched corpus and favour older filings; they are a signal of influence, not of current commercial weight.
Each row carries its publication number; clicking a row searches Eureka by that number.
Put your own technology through the same analysis
Eureka on the web
When you want the answer in the next five minutes.
The agent works the prompt against patents and technical literature, citing every source.
Run your analysis now →MCP server & REST API
When it has to run inside your own pipeline.
Patent search, landscape analysis and assignee resolution as MCP tools. Drop them into any agent framework, or call REST directly.
Browse MCP servers →What the data implies for a filing decision
Three patterns worth acting on before drafting or licensing in this space.
The claim space is held by a small group
Top five assignees account for 78.8% of all 66 records in scope, and the top ten reach 98.5%. A single leader holds 25 records outright. Freedom-to-operate work in this field should start with the leader's portfolio, not a broad prior-art sweep.
Activity peaked in 2022 and has since cooled
Filings rose to a peak of 15 in 2022 before falling to 1 by 2024, a -91% change over that span. Because publication lags filing by roughly 18 months, the 2025-26 figures are not yet a reliable read on current activity — the drop should be read as a real post-peak cooldown through 2024, not a live trend.
ML framing is present but still a minority
G01R and H01L together dominate the classification mix, reflecting the field's roots in classical electrical test and part-average statistics. G06N and G06T claims are present in roughly one in eight to one in eleven records, showing the machine-learning overlay on wafer-map and outlier work is real but not yet the default framing.
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to latent defect and outlier detection, with the prior art for and against each one.
Assignee landscape and where the gaps sit
The ranking is short and top-heavy — 19 companies cover the entire dataset, and most of the volume sits with a handful of them.
One assignee holds more than a third of the corpus alone
The top-ranked assignee's 25 records already exceed a third of all 66 records in scope. Combined with the next four assignees, the top five reach 78.8% of the field — a strong signal that core test-flow claims (electrical bin outlier detection, part-average screening) are already staked out.
A thin tail of smaller and adjacent entrants
Below the top ten — which together hold 98.5% of all records — the remaining ranked assignees hold only single-digit counts each. Several are not core semiconductor test vendors, pointing to defect-detection methods borrowed from adjacent domains such as network anomaly correlation and genomics assay screening.
Co-filing is limited and centred on one assignee
Only 10 co-assignee pairs appear across the dataset, and the strongest repeated pairings all involve the same mid-tier filer working with individual named inventors. This suggests most patenting here is done in-house rather than through joint ventures or foundry-fabless co-development.
| Assignee | Recent year | YoY |
|---|---|---|
| Ford Global Technologies, LLC | 1 | — |
| KLA Corporation | 0 | — |
| Essenlix Corporation | 0 | — |
| TEST ADVANTAGE | 0 | — |
| President and Fellows of Harvard College | 0 | — |
| AT&T Intellectual Property I, L.P. | 0 | — |
| Yahoo Assets LLC | 0 | — |
| International Business Machines Corporation (IBM) | 0 | — |
Where to take this analysis
The dataset points to a concentrated core and a thin edge — here is how to use both.
Map the leader's claim boundaries
With one assignee holding 25 of 66 records, a claim-by-claim read of that portfolio against your own screening method is the fastest way to find out whether you are inside or outside the occupied space.
Explore the leader's portfolio in Eureka →Test the white-space chips against your own R&D roadmap
Burn-in alternatives, spatial signature classifiers and cross-domain outlier correlation all show thin filing activity relative to the electrical-test core — worth checking against any screening method already in development.
Run a white-space search in Eureka →Common questions about this field
One assignee leads the ranked field with 25 of the 66 records in scope, well ahead of the rest of the ranking. The top five assignees combined hold 78.8% of all 66 records, and the top ten reach 98.5%, so the field is concentrated rather than fragmented. Any competitive or freedom-to-operate review should prioritise the leading assignee's portfolio before a broader prior-art sweep.
Filings rose to a peak of 15 in 2022, then fell to 1 by 2024, a -91% drop across that three-year span. That decline should not be extended into 2025 or 2026, because publication typically lags filing by roughly 18 months and those most recent years are still filling in. Treat 2024 as the most recent complete year for trend purposes.
Over half of the 66 records (53.0%) carry an IPC code in G01R, covering electric and magnetic measurement, and a quarter (25.8%) sit in H01L, semiconductor devices — the classical electrical-test core. Machine-learning framing shows up in G06N (12.1%) and image-based methods in G06T (9.1%), indicating an AI overlay on top of the statistical core rather than a replacement for it. A record can carry several IPC codes, so these shares add up to more than the total record count.
This KLA-assigned filing describes Z-direction Part Average Testing that correlates electrical test bin outliers with fab characterization data gathered during wafer fabrication. It ties statistical outlier flagging to physical process signals rather than treating electrical test data alone, which is a specific integration claim rather than a claim over outlier detection broadly. Anyone building a screening method that also cross-references fab process data against electrical bin results should read its claims closely before finalising a design.
The under-claimed branches sit outside the dense G01R/H01L core: burn-in alternative screening rules, spatial wafer-map signature classifiers, zero-defect target scoring for automotive-grade parts, and cross-domain outlier correlation methods borrowed from network or genomics-style anomaly detection. These areas show thin filing activity relative to the electrical-test core in this dataset, which points to open claim space rather than an absence of technical need.
Research Latent Defect and Outlier Detection in depth with Eureka
Go past this page: query the whole latent defect and outlier detection corpus yourself, in your own scope.
Every answer comes back with patent numbers you can open.
Disclaimer. This page is generated from Patsnap Eureka data drawn from a limited snapshot of global patent and scientific-literature records, and is provided for general information and reference only.
Patent data carries inherent limitations: recent filings (typically the most recent 18–24 months) are under-counted due to standard publication lag; counts may be reported at either a patent-family or a patent-record basis and are not always directly comparable; classification, applicant-name, and citation data may contain errors, duplicates, or omissions; and the underlying search query defines and constrains the scope shown. As a result, the analysis may be incomplete or inaccurate and may not reflect the full technology landscape.
Nothing on this page constitutes an exhaustive prior-art, novelty, freedom-to-operate, or validity search, nor does it constitute legal, financial, investment, or professional advice, and it should not be relied upon as such. Any patent, commercial, or strategic decision should be verified independently and reviewed with qualified patent, legal, and domain professionals. Patsnap makes no warranties, express or implied, as to the accuracy, completeness, or fitness for any particular purpose of the information presented.
Machine translation. Assignee and organisation names originally recorded in Chinese, Japanese or Korean have been rendered into English by an AI translation step so that the tables stay readable. These renderings are best-effort and may not match a company’s registered English name; the original name is what the underlying patent record carries, and it is what any Eureka query launched from this page uses.