Revisions to the Patent Examination Guidelines Effective as of January 1, 2026

Revisions to the Patent Examination Guidelines Effective as of January 1, 2026

Download PDF Version↓

Details of the Revised Patent Examination Guidelines

  1. Overview

On November 10, 2025, the China National Intellectual Property Administration (CNIPA) announced revisions to the Patent Examination Guidelines, which set out the operational rules and examination standards governing patents, utility models, and designs in China. The revised Guidelines will take effect on January 1, 2026.

A draft version of the revisions was released on April 30, 2025, followed by a public consultation. This article outlines the key contents of the final revisions and highlights the major changes to the draft version.

 

  1. Summary of Major Revisions

The revisions cover a broad range of areas, including formal examination, substantive examination, PCT-related practices, reexamination and invalidation proceedings, as well as other procedural matters. We summarize the revisions into the following 18 key points.

Among these changes, the refinement of examination standards and introduction of illustrative examples for artificial intelligence (AI)-related inventions and inventions involving bitstreams have attracted particular attention. In addition, the enhanced requirements for inventor information in application forms, the further clarification of the criteria for assessing inventive step, and certain modifications to the rules governing the calculation of Patent Term Adjustment (PTA) extension periods are expected to have a significant impact on patent prosecution practice.

(18) Partial revision of the calculation rules for PTA extension period

 

  1. Details of the Revisions
  • ●Formal Examination

(1) Stricter requirements for inventor eligibility

In response to the rapid development of artificial intelligence technologies, the revised Guidelines emphasize that inventors stated in a patent application form must be natural persons and must qualify as true inventors, defined as those who “have made creative contributions to the substantive features of the invention".

The revisions further clarify that inventor eligibility will generally not be examined during prosecution, unless there is evidence indicating that a listed inventor is not a true inventor.

(2) Expansion of inventor information required in the application form

Under the current practice, only the first inventor is required to provide identification information in the application form, namely nationality, and, for Chinese nationals, an ID card number. The revisions now require identification information to be provided for all inventors.

According to the explanations provided at a recent briefing session held by the CNIPA on the revisions, the underlying intent of the revision is to require the disclosure of identification information (including nationality and identification document number) for all inventors, including foreign inventors. However, based on our subsequent confirmation with CNIPA, for an initial period starting from January 1, 2026, only Chinese inventors will be required to provide both their nationality and ID card number, while foreign inventors will be required to provide nationality information only. Should there be any changes in CNIPA’s examination practice after the revised Guidelines come into effect, we will provide further updates in due course.

In addition, while the draft revisions released in April had provided that patent agencies would be responsible for the accuracy of inventor identification information, applicant identification information, and contact details included in the application form, the final revisions narrow this obligation. Under the revised Guidelines, patent agencies are required only to verify the authenticity of the applicant’s identification information and contact details.

 

(3) Clarification of practice regarding priority claims in divisional applications

The revised Guidelines clarify the handing of priority claims in divisional applications. Where a parent application claims priority but a divisional application derived from that parent does not claim priority at the time of filing, CNIPA will issue a notification stating that the priority claim is deemed not to have been made. The applicant may request restoration of the priority claim by paying the prescribed restoration fee within two months from the date of receipt of the notification.

In practice, even under the existing regime, priority restoration has generally been permitted where the parent application claimed priority, notwithstanding the absence of a priority claim in the divisional application at filing. Accordingly, this revision does not introduce a substantive change to current practice, but rather clarifies and formalizes the priority restoration procedure.

 

  • ●Substantive Examination

(4) Clarification of the definition of "plant varieties" excluded from patent protection

Article 25, Paragraph 1, Item (4) of the Patent Law provides that "animal and plant varieties" are excluded from patent protection. Under the revised Guidelines, the term "plant variety" is now defined as "a group of plants that have been artificially selected and bred or discovered and subsequently improved, exhibiting consistent morphological characteristics and biological characteristics and relatively stable genetic traits". This definition is consistent with those adopted in China’s Seed Law and the Regulations on the Protection of New Varieties of Plants.

On the one hand, the revision ensures alignment in terminology with the Seed Law and the Regulations on the Protection of New Varieties of Plants, such that plant varieties meeting this definition fall within the scope of plant variety rights protection as defined under the Seed Law. On the other hand, it clarifies that breeding intermediate materials and other plant-related subject matter that do not meet this definition may still qualify for protection under the Patent Law. This revision thus provides a clear delineation of the complementary relationship between the two legal frameworks.

The revised Guidelines also clarify that naturally occurring wild plants discovered in nature without any technical processing constitute "scientific discoveries" under Article 25, Paragraph 1, Item (1) of the Patent Law and are therefore not patentable. However, wild plants that have undergone artificial selection, breeding, or improvement and thereby acquired industrial applicability are no longer regarded as "scientific discoveries" and may constitute eligible subject matter for patent protection.

In addition, the CNIPA’s Interpretation of the Revisions to the Patent Examination Guidelines, published on December 4, 2025, provides more detailed examples regarding the determination of whether a subject matter constitutes a "plant variety". These examples are not discussed in detail here.

 

(5) Partial revision of the "same-day dual filing" system

Under the so-called "same-day dual filing" system, an applicant may file both an invention patent application and a utility model application for the same invention on the same day, provided that the applicant expressly declares its reliance on this system in each respective application form.

Under the current system, once the utility model application has been granted, and where the corresponding invention patent application is found during examination to satisfy all other requirements for grant, the applicant will be notified to choose either to abandon the utility model right or to amend the patent application to avoid double patenting. If the applicant elects to forgo the utility model right in response to such notification, the invention patent will be granted, and the utility model right will terminate as of the publication date of the grant of the invention patent.

By contrast, under the revision, first, it is clarified that where an applicant files both an invention patent application and a utility model application for the same invention on the same day but fails to declare reliance on the same-day dual filing system in the respective application forms, such applications will be handled under the ground for refusal due to double patenting pursuant to Article 9, Paragraph 1 of the Patent Law. Second, where the applicant has properly declared reliance on the same-day dual filing system at filing, and no grounds for rejection are found during the examination of the invention patent application, the applicant will be notified to declare, within a specified time limit, whether to abandon the utility model right. If the applicant forgoes the utility model right, the invention patent will be granted, and the utility model right will terminate as of the publication date of the grant of the invention patent. If the applicant does not agree to abandon the utility model right, the invention patent application will be rejected. In addition, where the applicant fails to respond within the specified time limit, the invention patent application will be deemed withdrawn.

According to explanations given at a recent briefing session on the revision held by the CNIPA on the revision, after the revision takes effect, where applications are filed under the same-day dual filing system, the patent application cannot be granted unless the previously granted utility model right is abandoned. This applies even if the claims of the invention patent application are amended to differ from those of the utility model. Moreover, this revised approach will apply to all applications which will be registered or after January 1, 2026, regardless of the filing date. At present, however, detailed operational practices remain unclear, and CNIPA’s actual practice following the implementation of the revisions remains to be observed.

In its interpretation, CNIPA explained that the same-day dual filing system was initially introduced to address examination backlogs and the lengthy time required to obtain patent grants. However, with the growing number of applications in high and new technology fields such as artificial intelligence, big data, and genetic technologies, use of this system has gradually declined. At the same time, the examination period for invention patent applications has been significantly shortened, and various examination acceleration mechanisms, such as prioritized examination, have been introduced, further reducing the practical need for reliance on this system. Against this background, the revisions to the Patent Examination Guidelines are intended to strengthen the procedural requirements of this system and to limit the applicants’ flexibility, so as to ensure that the system is used only in a manner consistent with its initial policy objectives.

 

(6) Addition of examination criteria and illustrative examples for assessing inventive step

The revised Guidelines introduce an explicit clarification in the inventive step examination criteria, stating that "features that do not contribute to the solution of a technical problem generally do not affect the assessment of the inventive step of an invention, even if they are written into the claims."

To illustrate the application of this principle, a specific example has been added.

【Example】

An invention relating to a camera addresses the technical problem of how to achieve more flexible shutter control, which is accomplished by improving the internal mechanical and circuit structures of the camera. After the examiner pointed out that the claims lacked inventive step, the applicant added features to the claims, including the shape of the camera housing, the size of the display screen, and the location of the battery compartment.

However, the specification does not explain any relation between these newly added features and the solution to the stated technical problem. These added features are either conventional components implied in the subject matter of the claims themselves, or they could be obtained by a person skilled in the art based on their ordinary technical knowledge and conventional experimental methods. The applicant has also failed to provide evidence or sufficient reason to demonstrate that these technical features bring any further technical effects to the claimed solution.

Accordingly, the aforementioned technical features do not contribute to the solution to the stated technical problem and do not bring inventive step to the claimed technical solution.

 While this revision largely reflects existing examination practice, it may still have a non-negligible impact on future assessments of inventive step.

Under China's approach to inventive step, recognition of inventiveness based on a particular technical feature generally requires that the feature produce an improved technical effect. This requirement is especially pronounced in the chemical field, where such effects are typically expected to be supported by experimental data or similar evidence. In this sense, even prior to the revision, examiners have focused on whether claimed features meaningfully contribute to solving the technical problem addressed by the invention. The revised Guidelines provide explicit textual support for this existing examination approach. Following the revision, examiners may exercise broader discretion in distinguishing between technical features that contribute to solving the technical problem of the invention and those that do not, and to assess inventive step based solely on the former.

After the revision, where an examiner determines that a specific technical feature recited in the claims does not contribute to solving the identified technical problem, applicants should first consider whether the technical problem identified by the examiner is appropriate and, if they believe the identification to be incorrect, raise arguments to challenge it. Applicants may also rebut the examiner’s conclusion by relying on disclosures in the specification or by submitting supplementary experimental data. As a result, under the revised practice, it will become even more critical for specifications to clearly explain the relationship between each technical feature and the technical effects it achieves.

Furthermore, the new provision may also serve as a basis for rejecting applications in which amendments introduce technical features unrelated to the problem solved by the invention. In practice, when responding to an examiner’s opinion citing a lack of inventive step, applicants often amend the claims by adding non-essential features in an effort to avoid an immediate rejection, while simultaneously challenging the examiner’s determination through written arguments. After the revision, however, the likelihood of such applications being directly rejected is expected to increase. Given CNIPA’s ongoing efforts to shorten examination timelines, this revision also reflects a clear intent to conserve examination resources and enhance examination efficiency.

 

(7) Addition of examination criteria and illustrative examples for AI-related inventions

One of the most notable aspects of this revision is the expansion of the examination standards for inventions related to artificial intelligence (AI) and similar technologies.

The revisions primarily include a change to the title of Part II, Chapter 9, Section 6 of the Examination Guidelines, from "Provisions on the Examination of Patent Applications Containing Algorithmic Features or Business Rules and Methods" to "Provisions on the Examination of Patent Applications Involving Artificial Intelligence, Big Data, etc., that Contain Algorithmic Features or Business Rules and Methods".

In addition, examination standards and illustrative examples relating to the requirements of public order and morality, inventiveness, and implementability have been further supplemented. Some of the newly added content follows the approach set out in the "Guidelines for Patent Applications for AI-Related Inventions (Trial)" published by the CNIPA in December 2024.

The specific revisions involve the following items 1) to 4).

 

1) Clarification of examination principles

Previously, it was stipulated that "the subject of examination for inventions containing algorithmic features or business rules and methods is the solution as recited in the claims". The revised Guidelines further add that "when necessary, the content of the specification shall also be examined". This change merely clarifies the existing examination principle.

 

2) Addition of examination criteria and examination examples for violations of public order and morality

New examination standards relating to public order and morality under Article 5, Paragraph 1 of the Patent Law have been introduced. Specifically, it is provided that “where an invention patent applications that contain algorithmic features or business rules and methods, in aspects such as data collection, tag management, rule setting, recommendation decision-making, etc., contain content that violates the laws, public order and morality, or harms the public interest, the patent right shall not be granted in accordance with Article 5, paragraph 1 of the Patent Law.”

In addition, two illustrative examples have been added as cases constituting violations of Article 5, Paragraph 1 of the Patent Law. It should be noted that, in Chinese patent examination practice, not only for AI-related inventions but all inventions that violate other existing Chinese laws and regulations, such as the Personal Information Protection Law, will be deemed to violate the public order and morality.

【Example 1】 A Big Data-Based System for Assisting in Selling Mattresses in a Mall

Application Content Overview

The invention patent application presents a big data-based system for assisting in selling mattresses in a mall. It collects customers’ facial feature information and obtains customers’ identification information through a camera module and a facial recognition module. The collected information is then analyzed to assess customers’ true preferences for mattresses, helping businesses with precision marketing.

Claims of the application

A big data-based system for assisting in selling mattresses in a mall, comprising a mattress display device and a management center, characterized in that:

the mattress display device includes a control module and an information acquisition module, configured to display and assist in the sale of mattress products and collect customer data; the control module is configured to interact with the management center; the information acquisition module includes a camera module and a face recognition module, configured to collect facial feature information of customers, adjust facial posture using a key point detection algorithm to obtain a normalized face image, locate a face region to be identified in the normalized face image using a face detection algorithm, and extract facial features within the face region using principal component analysis, to obtain the customer’s identification information;

the management center includes a management server and an analysis assistance system; the management server manages multiple mattress display devices; the analysis assistance system analyzes the data collected by the mattress display devices based on the customer’s identification information to obtain the customer’s true preferences and feeds back the analysis results to the management center.

Analysis and Conclusion

Relevant provisions of the Personal Information Protection Law of the People’s Republic of China stipulate that the installation of image collection and personal identification device in public places shall be carried out only for the purpose of maintaining public safety, comply with relevant national regulations, and be accompanied by clear and conspicuous notices. The collected personal images and identification information shall not be used for other purposes than maintaining public safety, except where individual consent is obtained.

It can be seen from the proposed solution that image capture and facial recognition is used for precision marketing of mattress in business premises such as shopping malls, which is not for the purpose of maintaining public safety. Furthermore, it is obvious that the collection of customers’ facial information and identification information to obtain and analyze their true preferences for mattress is conducted without their knowledge, and the application fails to demonstrate the legality or compliance of the data acquisition or information gathering. Therefore, this invention violates the law and, according to Article 5, Paragraph 1 of the Patent Law, cannot be granted a patent.

【Example 2】A Method for Building an Emergency Decision-Making Model for Autonomous Vehicles

Application Content Overview

The solution proposed in the invention patent application is a method for building an emergency decision-making model for autonomous vehicles. It uses the pedestrian’s gender and age as obstacle data, and uses the trained decision-making model to determine the protected object and the object to be collided with when it is impossible to avoid the obstacle.

Claims of the application

A method for building an emergency decision-making model for an autonomous vehicle, comprising:

acquiring historical environmental data and historical obstacle data for the autonomous vehicle, the historical environmental data including the vehicle’s speed, distance to obstacles in its lane, distance to obstacles in adjacent lanes, speed and direction of movement of obstacles in its lane, and speed and direction of movement of obstacles in adjacent lanes, and the historical obstacle data including the gender and age of  pedestrians,

performing feature extraction on the historical environmental data and historical obstacle data, which are used as input data for the decision-making model, and training the decision-making model based on the historical data, where the historical driving trajectory of the vehicle when it is unable to avoid the obstacle is used as the output data of the decision-making model, the decision-making model being a deep learning model,

acquiring real-time environmental data and obstacle data and determining, based on the trained decision-making model, the driving trajectory of the autonomous vehicle when it is unable to avoid the obstacle.

Analysis and Conclusion

This invention relates to a method for building an emergency decision-making model for autonomous vehicles. Human life has equal value and dignity, regardless of age or gender. If an emergency decision-making model for autonomous vehicle, in the case of unavoidable accidents, selects between the protected object and the object to be collided with based on the pedestrian’s gender and age, this contradicts the public’s ethical and moral concept that all lives are equal. Furthermore, this decision-making method reinforces existing gender and age biases in society, raises public concerns about public transportation safety, and undermines public trust in technology and social order. Therefore, this invention contains content that violates social morality and, according to Article 5, Paragraph 1 of the Patent Law, cannot be granted a patent.

3) Addition of illustrative examples for inventive step

Two illustrative examples have been introduced as part of the inventiveness examination for inventions related to artificial intelligence and similar technologies. The inventiveness examination standards themselves remain unchanged; the revision consists solely of the addition of these examples.

Both of the newly added examples involve the application of artificial intelligence algorithms or models to specific technical fields. In other words, they provided guidance on how inventive step is assessed in situations where the application scenarios differ from that of the cited prior art, while the underlying algorithms or models remains the same. In its Interpretation of the Revisions to the Patent Examination Guidelines, the CNIPA explains that “where the algorithm or model of the claimed invention differs from the prior art only in terms of the application scenario or processing object, but no substantive modifications have been made to the algorithmic flow, model parameters, or similar aspects, such an invention will generally be considered to lack inventiveness.”

In Example 18 titled "A Method for Identifying the Number of Ships from Images", when compared with a prior-art method for identifying the number of fruits on a tree disclosed in a cited reference, steps such as image information marking, dataset classification, and model training are not substantially changed. As a result, the method is deemed not to solve any technical problem specific to the field of ships, and is therefore determined to lack inventiveness.

Example 19, titled "A Method for establishing a neural network model for classifying scrap steel grades" is based on an actual examination decision. Although the application scenario of the method in Example 19—namely, the classification of scrap steel—is similar to that of the cited prior art, the claimed method addresses a different technical problem by adjusting the number of paths and level settings of convolutional and pooling layers during the model training process, thereby achieving effects distinct from those of the prior art. Accordingly, the invention is considered to have inventiveness, as it modifies the artificial intelligence algorithm or model to address technical problems in a specific application field and achieves beneficial technical effects. 

【Example 18】A Method for identifying the number of vessels

Application Content Overview

The invention patent application proposes a method for identifying the number of vessels, which acquires vessel image data and trains a detection data model through deep learning to solve the technical problem of accurately identifying the number of vessels in the current sea area.

Claims of the application

A method for identifying the number of vessels, comprising:

obtaining a dataset of vessel images, preprocessing the image information in the dataset, labeling position and boundary information of the vessels in the images, and dividing the dataset into a training dataset and a test dataset;

preforming deep learning using the training dataset to build a training model;

inputting the test data into the training model to obtain vessel test result data; and

determining actual number of vessels by multiplying the vessel test result data with a preset error parameter.

Analysis and Conclusion

Prior art document 1 discloses a method for identifying the number of fruits on a tree, and specifically discloses the steps of acquiring image information, labeling the position and boundaries of fruits in the image, dividing the dataset, training the model, and determining the actual number of fruits.

The only difference between the solution in the patent application and prior art document 1 lies in the objects to be identified. Although the vessels and fruits themselves differ in appearance, size, and environment, for those skilled in the art, the steps required to identify the actual quantity-such as information labeling, dataset dividing, and model training-all are performed with respect to the positional relationships of the objects to be identified in the image. The claims do not demonstrate any changes to the training methods or model levels during deep learning or model training due to the different objects being identified. There is no adjustments or improvements to the deep learning, model construction, or training process in labeling vessel data in the image as compared to labeling fruit data in the image to obtain a training dataset and then train the model. Therefore, the claimed invention lacks inventiveness.

 

【Example 19】A Method for building a neural network model for  scrap steel grading

Application Content Overview

Scrap steel needs to be graded according to its average size during collection and storage. However, it is often disorganized and piled up during collection and storage, making manual size measurement and grading inefficient and inaccurate. This invention patent application proposes a method for building a neural network model for scrap steel grading. By using a convolutional neural network to learn and form a neural network model for grading with grades as outputs, the efficiency and accuracy of scrap steel grading can be improved.

Claims of the application

A method for building a neural network model for scrap steel grading, the model being used for grading  the stored scrap steel, the method including:

acquiring multiple images, determining the different scrap steel grades of multiple images, preprocessing the images, extracting data features from the images with different grades, and performing convolutional neural network learning on the extracted data features of the images with different grades to form a neural network model for grading with grades as output, wherein,

the extraction of the image data features is to extract the set calculated by performing convolutional neural network convolution on the pixel point matrix data of the image screen, including the extraction of the color, edge features and texture features of the object in the image, realized by a set of outputs of a plurality of lines consisting of the convolutional layer or convolutional layer plus pooling layer, and the extraction of the association features between the edge and texture of the object in the image;

wherein the extraction of the color and edge features of the object in the image is realized by the set of outputs of three lines consisting of convolutional layer plus pooling layer, including, from left to right, a first line consisting of one pooling layer, a second line  consisting of two convolutional layers and a third line consisting of four convolutional layers; and the extraction of texture features in the image is realized, after making the set of the extraction results of the object color and edge feature, by the set of outputs of three lines consisting of convolutional layer, including, from left to right, a first line consisting of a convolutional layer of zero convolution, a second line consisting of two convolutional layers, and a third line consisting of three convolutional layers; and

the number of lines for calculations in the convolutional layer for extracting association features between edges and textures is greater than the number of lines for calculations in the convolutional layer for extracting the color, edge and texture features of the object in the image.

Analysis and Conclusion

To address the challenges of accurately identifying scrap steel as a type of raw material, stamping waste, bread iron, or other materials due to the complexity, variety, and material differences of recycled resources, and to improve the recycling rate of recycled resources, Prior Art Document 1 provides a method for identifying the type of scrap steel based on a convolutional neural network model. Specifically, it discloses the steps of acquiring image data of multiple scrap steel with determined types, preprocessing the image data for feature extraction, and training the convolutional neural network to obtain a product model.

The difference between the solution in the invention patent application and prior art document 1 lies in the different training data and extracted features, as well as the number of lines and level settings of convolutional and pooling layers. Compared to prior art document 1, the technical problem to be actually addressed by the invention is how to improve the accuracy of scrap steel grading. Prior art document 1 uses image data of scrap steel with determined types for feature extraction and model training. The invention patent application, in order to grade scrap steel based on its average size, needs to identify the shape and thickness of the scrap steel from the images of scrape steel that are chaotic and overlapping. To extract features such as color, edges, and texture of scrape steel from the images, the number of lines and level settings of convolutional and pooling layers were adjusted during model training. These algorithmic and technical features are mutually supportive and interactive, improving the accuracy of scrap steel grading. The contribution of these algorithmic features to the technical solution should be considered. The aforementioned adjustments to the number of lines and level settings of convolutional and pooling layers have not been disclosed in other prior art documents, nor are they common knowledge in the field. The prior art as a whole does not provide any inspiration for improving the aforementioned prior art document 1 to obtain the present invention patent application, and the claimed invention technology solution possesses inventiveness.

 4) Addition of examination criteria and examination examples related to implementability

With respect to the disclosure requirements for drafting specifications for inventions related to artificial intelligence and similar technologies, the revised Guidelines provides that, if an invention involves the construction or training of an AI model, the specification generally needs to clearly describe the necessary modules, the hierarchical structure or connections of the model, as well as the specific steps and parameters required for training.

Furthermore, for inventions in which an artificial intelligence model or algorithm is applied to a specific technical field or application scenario, the specification generally needs to clearly describe how the model or algorithm is integrated with that technical field or application scenario, and how the input and output data of the algorithm or model are set to demonstrate their intrinsic connection, so that a person skilled in the art can implement the invention based on the content disclosed in the specification.

The provision aims to address the so-called “black box” problem in the specifications for AI-related inventions.

In addition, two new illustrative examples have been added in this regard.

【Example 20】 A method for generating facial features

Application Content Overview

The invention patent application achieves information sharing among the second convolutional neural networks by sharing the feature region image set generated by the first convolutional neural network with a spatial transformation network. This reduces memory resource consumption and improves the accuracy of face image generation results.

Claims of the application

A method for generating facial features, including:

acquiring an image of a face to be identified;

inputting the face image to be identified into a first convolutional neural network to generate a set of feature region images of the face image to be identified, wherein the first convolutional neural network is used to extract feature region images from the face image;

inputting each feature region image in the feature region image set into the corresponding second convolutional neural networks to generate regional face features of the feature region image, wherein the second convolutional neural network is used to extract the regional face features of the corresponding feature region image; and

generating a set of facial features of the face image to be identified based on the regional facial features of each feature region image in the feature region image set; wherein,

the first convolutional neural network also includes a spatial transformation network for determining the feature regions of the face image; and

the inputting the face image to be identified into a first convolutional neural network to generate a set of feature region images of the face image to be identified includes: inputting the face image to be identified into the spatial transformation network to determine the feature regions of the face image to be identified; and

inputting the face image to be identified into the first convolutional neural network to generate a set of feature region images of the face image to be identified based on the determined feature regions.

Relevant paragraphs of the specification

The method for generating facial features provided in this application firstly generates a set of feature region images of the face image to be recognized by inputting the acquired face image to be recognized into a first convolutional neural network. The first convolutional neural network can be used to extract feature region images from the face image. Then, each feature region image in the feature region image set can be input into a corresponding second convolutional neural network to generate the regional face features of that feature region image. The second convolutional neural network can be used to extract the regional face features of the corresponding feature region image. Subsequently, based on the regional face features of each feature region image in the feature region image set, a set of facial features of the face image to be recognized can be generated. In other words, the set of feature region images generated by the first convolutional neural network can share information among the various second convolutional neural networks. This reduces the amount of data, thereby reducing memory resource consumption and improving generation efficiency.

To improve the accuracy of the generated results, a spatial transformation network can also be included in the first convolutional neural network to determine the feature regions of the face image. In this case, the electronic device can input the face image to be recognized into the spatial transformation network to determine its feature regions. Then, for the input face image to be recognized, the first convolutional neural network can extract images matching the feature regions on the feature layer based on the feature regions determined by the spatial transformation network, thereby generating a set of feature region images of the face image to be recognized. The specific location of the spatial transformation network within the first convolutional neural network is not limited in this application. The spatial transformation network can continuously learn to determine the feature regions of different features in different face images.

Analysis and Conclusion

The invention patent application relates to a method for generating facial features. In order to improve the accuracy of the facial image generation results, a spatial transformation network can be provided in the first convolutional neural network to determine the feature regions of the facial image. However, the specification does not describe the specific position of the spatial transformation network in the first convolutional neural network.

Those skilled in the art will understand that the spatial transformation network, as a whole, can be inserted into any position within a first convolutional neural network (CNN) to form a nested CNN structure. For example, the spatial transformation network can serve as the first layer or an intermediate layer of the first CNN, without affecting its ability to identify feature regions of an image. Through training, the spatial transformation network can determine the feature regions containing different features of different face images. Therefore, the spatial transformation network can not only guide the first CNN to segment feature regions but also perform simple spatial transformations on the input data to improve the processing performance of the first CNN. Accordingly, the hierarchical structure of the model used in the invention patent application is clear, and the input/output relationships between each layer are also clear. Both the CNN and the spatial transformation network are well-known algorithms, and those skilled in the art can construct the corresponding model architecture based on the above description. Therefore, the solution claimed in the invention patent application has been fully disclosed in the specification and complies with Article 26, Paragraph 3 of the Patent Law.

 

【Example 21】A method for predicting cancer based on bioinformation

Application Content Overview

The invention patent application provides a method for predicting cancer based on bioinformation. By using a trained enhanced screening model for malignant tumor, complete blood count, blood biochemical test indicators and facial image features are used as inputs to the screening model to obtain a prediction value for malignant tumor disease, thereby solving the technical problem of improving the accuracy of malignant tumor prediction.

Claims of the application

A method for predicting cancer based on bioinformation, comprising:

obtaining the complete blood count and blood biochemistry test reports of a subject to be screened, and identifying the test indicators, age, and gender in the complete blood count and blood biochemistry test reports;

obtaining a frontal, makeup-free facial image of the subject to be screened and extracting facial image features; and

calculating, based on the enhanced screening model for malignant tumors, a predicted value of malignant tumor incidence for the corresponding subject to be screened, wherein,

the training process of the enhanced screening model for malignant tumors includes:

constructing a large-scale population sample set, which includes the complete blood count, blood biochemistry and facial images of the same subject;

creating learning samples using the features of the complete blood count, blood biochemistry and facial images; and

training machine learning algorithm model using the learning samples to obtain the enhanced screening model for malignant tumors.

Relevant paragraphs of the specification

Currently, when using tumor markers to identify malignant tumors, a tumor marker level above a threshold cannot definitively confirm a malignant tumor, while a level below the threshold does not rule out a malignant tumor. Therefore, the accuracy of predicting cancer based on tumor markers is low. This application utilizes complete blood count (CBC), blood biochemistry indicators, and facial images to improve the accuracy of identifying various malignant tumors. This application, while utilizing blood test data, also considers the health status of the subject being screened as reflected in facial images, enabling a more accurate prediction of the probability of malignant tumors incidence. For the selection of features for calculation in the enhanced screening model for malignant tumor, some or all of the indicators from CBC and blood biochemistry may be utilized.

Analysis and Conclusion

The technical problem to be solved by this invention patent application is how to improve the accuracy of malignant tumor prediction. To address this problem, the solution utilizes a pre-trained enhanced screening model for malignant tumor, taking complete blood count, blood biochemistry indicators, and facial image features as inputs, to obtain a predicted value for malignant tumor incidence. However, both complete blood count and blood biochemistry tests, as common biochemical test items, each contain dozens of indicators. The specification does not specify which indicators are key indicators related to tumor prediction accuracy, nor does it clarify whether all indicators are used and different weights are assigned to various indicators for prediction. Those skilled in the art cannot determine which indicators can be used to diagnose malignant tumors. Furthermore, based on current scientific research, it remains uncertain whether there is a correlation between facial features and the occurrence of malignant tumor, except for a few types of tumors such as facial skin cancer. The specification also does not describe or prove a causal relationship between the “basis factors for judgment” and the “judgment result.” In addition, the specification does not provide any validation data to prove that the accuracy of identifying various malignant tumors using this solution is higher than that using tumor markers, or is significantly higher than the accuracy of randomly judging the probability of malignant tumor incidence. Based solely on the disclosure in the specification, a person skilled in the art would be unable to determine that the solution in this application can solve the technical problem it seeks to address. Therefore, the technical solution for which protection is sought in the invention patent application is not fully disclosed in the specification, and the specification does not comply with Article 26, Paragraph 3 of the Patent Law.

(8) Addition of examination criteria and illustrative examples for inventions involving bitstreams

In this revision, following Part II, Chapter 9, Section 6 titled “Provisions on the Examination of Patent Applications Involving Artificial Intelligence, Big Data, etc., that Contain Algorithmic Features or Business Rules and Methods” as introduced in item (7) above—a new Section 7, “Examination Criteria for Inventions Involving Bitstreams” has been added.

1) Addition of examination criteria relating to patentable subject matter

Firstly, with respect to patentable subject matter, it is stipulated that claims directed merely to a simple bitstream, or claims in which substantially all content other than the subject matter of the claim refers solely to a simple bitstream, fall under “rules and methods for intellectual activities” as defined in Article 25, Paragraph 1, Item (2) of the Patent Law, and therefore are not eligible for patent protection. Examples provided include claims such as "A bitstream characterized by comprising syntax element A, syntax element B, ..." or "A method for generating a bitstream characterized by comprising syntax element A, syntax element B, ...".

In the technical field of digital video encoding and decoding, if a video encoding/decoding method for generating a specific bitstream falls under an "invention" as stipulated in Article 2, Paragraph 2 of the Patent Law, then the method for storing or transmitting the bitstream, as defined by that encoding/decoding method, or a computer-readable storge medium for storing it, may achieve optimal allocation of storage or transmission resources. Accordingly, the storage or transmission method and the computer-readable storage medium also fall within the scope of "invention" under Article 2, Paragraph 2 of the Patent Law and are eligible for patent protection.

2) Addition of examination criteria relating to implementability

It is stipulated that, for patent applications involving a bitstream generated by a specific video encoding/decoding method, the specification should describe the specific video encoding/decoding method clearly and completely so that a person skilled in the art to implement it.

In addition, where protection is sought for methods of storing or transmitting the bitstream, or for computer-readable storage media storing the bitstream, corresponding explanations must be provided in the specification.

3) Addition of examination criteria for claim drafting

It is stipulated that in patent applications involving a bitstream generated by a specific video encoding/decoding method, claims may be drafted in the form of method claims, apparatus claims, and computer-readable storage medium claims. In a single application, such claims should generally be drafted based on claims directed to the specific video encoding method used to generate the bitstream, and should be formulated by referencing such encoding method claims or by including all of their technical features.

The revision provides specific examples of acceptable claim formats.

【Example 1】

1. A video encoding method, comprising:

frame division step, …

entropy coding step,…

2. A video encoding apparatus, comprising:

a frame division unit, …

an entropy coding unit, …

3. A video decoding method, comprising:

entropy decoding step,…

frame output step, …

4. A video decoding apparatus, comprising:

an entropy decoding unit, …

a frame output unit, …

5. A method for storing a bitstream, including:

performing the video encoding method of claim 1 to generate the bitstream; and

storing the bitstream.

6. A method for transmitting a bitstream, including:

performing the video encoding method of claim 1 to generate the bitstream; and

transmitting the bitstream.

7. A computer-readable storage medium having a computer program/instructions and a bitstream stored thereon, characterized in that the computer program/instructions, when executed by a processor, implement the video encoding method of claim 1 to generate the bitstream.
  • ●PCT Application Related

(9) Clarification of the signatory for priority assignment documents upon national phase entry

With respect to priority claims made in PCT applications, the current Examination Guidelines provide that where the applicant of a PCT application is not included among the applicants of the priority application, but has obtained the priority right through assignment, gift or the other means from the applicant(s) of the priority application, the PCT applicant must submit to the CNIPA documentary proof signed/sealed by the "assignor".

The recent revision changes the signatory from the "assignor" to "all applicants of the priority application". This revision merely aligns the wording with other parts of the examination guidelines and does not entail any substantive change in practical procedures.

 

  • ●Reexamination and Invalidation Proceedings

(10) Simplification and omission of the composition of examination decisions

For examination decisions in reexamination proceedings (appeals against rejection) and invalidation proceedings, the current Examination Guidelines require inclusion of the following components: 1) bibliographic data, 2) legal basis, 3) main points of the decision, 4) brief of the case, 5) grounds of decision, 6) conclusion, and 7) drawings. For decisions revoking a rejection in reexamination proceedings, component 4) "brief of the case" may be simplified or omitted.

Whereas, the revised Guidelines stipulate that examination decisions shall "generally include" components 1) to 7), thereby granting the adjudication panel greater discretion in drafting the content of the decision. Correspondingly, the provision allowing simplification or omission of the "brief of the case" in reexamination decisions revoking rejections has been deleted. This revision is generally not expected to result in substantive changes in practice.

 

(11) Stricter qualification requirements for invalidation requesters

A new ground for non-acceptance of invalidation requests has been added: "the invalidation request is not made based on the true intention of the requester". The Interpretation of the Revisions to the Patent Examination Guidelines pointed out that: In practice, cases have arisen where invalidation requests were filed under another person’s name. In such cases, the invalidation request is not made based on the true intention of the requester and is often accompanied by the submission of forged signatures, forged power of attorney, and other related documents. Such conduct violates the principle of good faith and undermines both the credibility of the patent invalidation system and the order of market competition.

Prior to this revision, there had already been a practical trend toward stricter scrutiny of qualifications of invalidation requesters. In recent CNIPA practice, when an invalidation request is filed in the name of a natural person, a notification may be issued requesting verification of the requester’s identity information and confirmation that the invalidation request represents his or her true intention. To comply, the requester must either appear in person at the CNIPA with valid identification documents, or submit notarized documents providing that the notary has verified the requester's identity, certified that the requester has confirmed the invalidation request reflects his or her true intention, and witnessed the requester’s signing of the declaration form attached to the aforementioned notification. If neither procedure is completed within 15 days from receipt of the notification, the invalidation request will be deemed withdrawn.

Shortly after the publication of the revised Guidelines, on November 15, 2025, the CNIPA issued an invalidity trial decision in a case concerning the qualification of an invalidation requester, attracting attention within the patent community (Decision No. 4W119542). In that case, the patentee argued that the invalidation requester, born in 1949 and residing in Taiwan, lacked any academic or professional experience in pharmaceuticals or patent-related fields, yet had repeatedly filed invalidation requests against pharmaceutical-related patents. On this basis, the patentee contended that the invalidation request did not reflect the requester’s true intention and should therefore be deemed invalid. The invalidation requester submitted a notarized affidavit stating that the request was a true expression of intent. However, the patentee submitted a handwriting verification report indicating a high probability that the signature on the affidavit and the signature on the power of attorney filed with the request for trial were made by different persons. Consequently, the examination panel concluded that the invalidation request, having been based on forged legal documents, was invalid and should not be accepted.

According to the Interpretation of the Revisions to the Patent Examination Guidelines, the CNIPA is considered to take a critical view of invalidation requests filed by so-called "straw man" as illustrated by the above decision. While this does not mean that invalidation requests filed in such a manner have become entirely impossible, it is expected that scrutiny of requester eligibility will become increasingly stringent going forward. 

 

(12) Clarification of the scope of the "res judicata" principle for invalidation grounds

The revised Guidelines stipulate that not only invalidation requests based on grounds that are "the same" as those on which an examination decision has already been rendered will not be accepted pursuant to the principle of res judicata, but invalidation requests filed based on "substantially the same grounds" will likewise not be accepted or examined.

The Explanation of the Revised Patent Examination Guidelines provides the following two specific examples to illustrate how “substantially the same grounds” are to be determined.

Example 1:
Grounds in the earlier invalidation: Feature B of claim 1 broadly encompasses multiple embodiments; however, the specification describes only one of those embodiments, and therefore Claim 1 fails to satisfy the support requirement.

Decision of the examination panel: claim 1 satisfies the support requirement.

Grounds in the later invalidation: Feature B of claim 1 includes functional limitations, and a person skilled in the art cannot understand that the function may also be achieved by other alternative means not described in the specification; therefore, Claim 1 fails to satisfy the support requirement.

Example 2:
Grounds in the earlier invalidation: Claim 1 lacks inventive step over Evidence 1 in combination with common general knowledge.

Decision of the examination panel: claim 1 has an inventive step.
Grounds in the later invalidation: Claim 1 lacks novelty over Evidence 1.

In both examples above, when the examination decision is rendered on the earlier invalidation grounds, the determination of the later invalidation grounds has already been made clear. Accordingly, such later grounds are deemed to constitute "substantially the same grounds".

This revision aims to curb the practice of repeatedly filing invalidation requests based on substantially the same grounds as a means of prolonging disputes with patentees. It also serves the purpose of conserving examination resources and improving procedural efficiency.

 

(13) Clearer rules on amendments during invalidation proceedings

The revised Guidelines expressly stipulates that when amendments are made during invalidation proceedings, the party must submit full replacement pages together with a comparison table showing the amendments.

In addition, it is stipulated that where multiple amended texts are submitted during the same invalidation proceeding and all such texts comply with the amendment requirements, only the last-submitted amended text should prevail, and other amended texts will not serve as the basis for examination.

 

  • ●Procedural Matters

(14) Conditional abolition of page-count surcharges for sequence listings

It is stipulated that sequence listings submitted in the prescribed electronic data format shall not be counted toward the number of the specification pages and shall not be subject to additional page-based filing fees. It should be noted, however, that sequence listings submitted in paper form will continue to be subject to the original rules for calculating additional fees.

In line with this change, the list of official fees applicable to the national phase of PCT applications has removed the provision stating that "where a nucleic acid sequence and/or amino acid sequence listing, as an independent part of the specification, exceeds 400 pages, the sequence listing shall be counted as 400 pages".

 

(15) Revision of the rules on requests for refund of official fees

For the following situations 1)~3), which under the current practice allow the CNIPA to proactively refund fees, refunds will, under the revised rules, be made only upon request by the relevant party.

The stated reason for this change is to ensure accuracy and timeliness in fee refunds and to protect the interests of the parties. Accordingly, it should be noted that, after the implementation of the new rules, if no refund request is filed by the party in situations 1)~3), the official fees will not be refunded.

1) Where a patent application is deemed withdrawn, a divisional application is deemed withdrawn, or a request for withdrawal of a patent application is approved before the Patent Office issues a notification that the invention patent application has entered the substantive examination stage, the party may request a refund of the substantive examination fee already paid.

2) Where annuity fees are paid after the announcement of a decision terminating the patent right or declaring the patent right invalid in its entirety, the party may request a refund of the annuity fees already paid.

3) Where a restoration of rights procedure is initiated and a decision rejecting the request is issued, the party may request a refund of the restoration request fee and related fees already paid.

 

(16) Explicit codification of expedited examination

It is stipulated that, upon applicant's request, an application may be subject to prioritized examination, expedited examination, or deferred examination.

It is further stipulated that patent applications submitted after passing preliminary examination by an Intellectual Property Protection Center or a Rapid Rights Protection Center may be subject to expedited examination, provided that the relevant requirements are satisfied.

For applications filed by domestic Chinese applicants, prioritized examination and accelerated examination have been widely used in practice, and this revision formally incorporates such practices into the rules.

(17) Clarification of the information shown on registration certificates for PCT applications

It is clarified that, for international applications and their divisional applications, the "inventor(s)/designer(s)/applicant(s) at the time of filing" listed on the registration certificates refer to the inventor(s)/designer(s)/applicant(s) at the time of the international application enters the Chinese national phase or, in the case of a divisional application, at the date of submission of the divisional application.

 

(18) Partial revision of the calculation rules for PTA extension period

Article 78, paragraph 3, item (1) of the Implementing Regulations of the Patent Law provides that "delay caused by reexamination procedure where patent application documents are granted after being amended in accordance the provisions of Rule 66 of these Implementing Regulations" are "reasonable delays" and therefore are not eligible for patent term compensation.

This revision further clarifies that even if no amendments are made during the reexamination process, the period spent on reexamination will likewise not be included in the compensable patent term if the prior rejection is overturned based on new grounds or evidence submitted by the requester after receiving the decision of refusal.

 

  1. Conclusion

This revision was finalized within a relatively short period following the public consultation on the draft revision conducted through June 15, and the changes from the draft version were relatively limited in scope. While many of the revisions are confirmatory in nature and are not expected to cause significant practical changes, certain modifications—such as the change to the same-day dual filing system for patents and utility models, and the stricter eligibility requirements for invalidation requesters—may have uncertain practical implications. The impact of these changes will require observation of how the CNIPA handles related matters in practice. My office will continue to monitor developments and will provide updates as soon as clear progress is observed.

Contact Us

For inquiries or consultations,
please contact us via the web form.

Contact Us by Email