DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Priority
Acknowledgment is made of applicant’s claim for foreign priority under 35 U.S.C. 119 (a)-(d). The certified copy has been filed in parent Application No. KR10-2024-0051402, filed on 04/17/2024.
Application Status
This office action is issued in response to application filed 10/21/2024. Claims 1-16 are pending. Claims 1-16 are rejected. This action is non-final. A three-month Shortened Statutory Period for Response has been set.
Claim Objections
Claim 1 is objected to because of the following informality:
Claim 1: The limitation “previous steam of inputs” includes the misspelling of the word “stream”.
Appropriate correction is required.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-16 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claims 1-16 recite the phrases "primary response" and/or “secondary response”. While it is clear that these “responses” are associated with the output of different language models, it is unclear what the “the metes & bounds” of these responses are; particularly with respect the “secondary response” controlling operation of a vehicle. The examiner recommends additional claim language to clearly define the metes & bounds of these responses.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-5 and 9-13 are rejected under 35 U.S.C. §103 as being unpatentable over the combination of Fernandez et al. (US 2025/0249742 A1, henceforth Fernandez) and Wang et al. (US 2025/0077895 A1, henceforth Wang).
Regarding Claim 1, Fernandez discloses the limitations: a method for controlling operation of a vehicle {input query 205 leads to operation of car control unit 251, Fig. 2}, the method comprising: acquiring, based on a first machine learning model {220, Fig. 2} associated with a current input and a previous stream of inputs {“two example input queries are considered in the following. A first example input query is “Turn on the wiper!”. A second example input query is “Route me to Paris, turn on the wiper, and provide the weather forecast there!”.”, ¶[0067]; also “a previous input query, a previous conversation”, ¶[0054]}, a primary response to the current input {output from first language model 220 into second language model 241, Fig. 2}; acquiring, based on applications of a second machine learning model {241, Fig. 2} and a third machine learning model {additional language model in the form of an external language model 243, Fig. 2; and “The plurality of second language models comprises a local second language model 241 and an external second language model 243.”, ¶[0073]} to the primary response, a secondary response {output from language model associated with processing unit 240, Fig. 2}, wherein the second machine learning model is tuned to provision of position information and weather information associated with the vehicle {the second input port to processing unit 240, Fig. 2, can provide additional parametric data to language model 241 via transceiver 242, such as the data types described in ¶[0054] and ¶[0069]}, and wherein the third machine learning model {243, Fig. 2} is tuned to provision of vehicle information {“The plurality of applications may comprise a routing application, a location search application, an electrical vehicle charger search application, a traffic application, a vehicle control application, a weather application, a web search application, a general knowledge application, a feedback application, a car manual application”, ¶[0069]}; outputting the adjusted secondary response, wherein the current input or the previous steam of inputs is related to a request for information of a destination area or a target area for the vehicle {“example input query is “Route me to Paris, turn on the wiper, and provide the weather forecast there!”, ¶[0067]}; and controlling, based on the adjusted secondary response, operation of the vehicle {input query 205 begins the process leading to an output from second language model (within processing unit 240) to car control unit 251, Fig. 2}.
Fernandez does not appear to explicitly disclose: adjusting, based on a fourth machine learning model the secondary response, wherein the fourth machine learning model is tuned to a length adjustment of the secondary response or tuned to verification of information associated with the secondary response.
However, Wang explicitly recites the limitation: adjusting, based on a fourth machine learning model the secondary response, wherein the fourth machine learning model is tuned to a length adjustment of the secondary response or tuned to verification of information associated with the secondary response {the use of multiple language models combined with fine-tuning, providing the adjusting: “Configuring the language model neural networks and performing a machine learning task can include leveraging the ability of a first large language model to follow prompt-engineered instructions and perform chain-of-thought reasoning, while also fine-tuning a second, smaller language model neural network to optimize the machine learning task performance”, Abstract; see also Fig. 1, which a second prompt and the output of the first language model are combined for inputting into a second language model, ¶[0011]}.
Fernandez and Wang are analogous art because both deal with using multiple language models for improving the accuracy of the final output.
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of Fernandez and Wang before them, to modify the teachings of Fernandez to include the teachings of Wang to improve the final output through the use of multiple language models and the use additional data input between models {“the intermediate output improves the performance of the downstream machine learning task performed by the second language model because the intermediate output helps direct how the second language model generates the final output through its inclusion in the second prompt.”, ¶[0090]}.
Regarding Claim 2, the combination of Fernandez and Wang discloses all the limitations of the method of Claim 1, as discussed supra. In addition, Fernandez explicitly recites the limitation: wherein the acquiring the primary response {output from first language model 220, Fig. 2} comprises: inputting {input query 205, Fig. 2} a first input for semantic inference to the first machine learning model {“a context used for evaluating the input query may be different from a context used for processing the input query. In particular, a context may comprise one or more of: a location, a previous input query, a previous conversation, a type of a car, a specification of a car, and a battery state of a car.”, ¶[0054]};
inputting a second input for function classification to the first machine learning model {“In a preferred embodiment, the first language model may perform a classification of the input query.”, ¶[0036]}; inputting a third input for a query creation to the first machine learning model {“evaluating the input query may depend, at least in part, on a location and/or a previous input query and/or a previous conversation and/or a type of a car and/or a specification of a car”, ¶[0042]}; searching, based on the query generated by the first machine learning model, for a document in a database {evaluation of input query requires identification of the vehicle applications, Fig. 3, involved in addressing the input query via apparatus 200 in Fig. 2, which includes “a web search application” (¶[0042]) and other applications requiring searching any onboard databases, look-up tables, etc.}; and inputting, based on content of the document, a fourth input for generation of the primary response {output from first language model 220, Fig. 2} to the first machine learning model {inputting multiple pieces of data for evaluation by a machine learning model, such a neural network, is well known in the art; additionally, the apparatus of Fernandez in Fig. 2 provides the appropriate output for all the vehicle applications in Fig. 3, which includes vehicle control signals, signals for operating vehicle systems (e.g., wipers 252, sunroof 255 and lights 256), and applications requiring audio, textual or graphical outputs (e.g., infotainment system 261, general knowledge 263 and web search 269.}.
Regarding Claim 3, the combination of Fernandez and Wang discloses all the limitations of the method of Claim 2, as discussed supra. In addition, Fernandez explicitly recites the limitation: wherein the acquiring the secondary response comprises: acquiring, based on positions of the destination and the target area, an estimated travel time from the destination to the target area; and adding the estimated travel time to the primary response {the vehicle applications includes a vehicle navigation application (“the input query comprises selecting one or more applications from a plurality of applications, wherein the plurality of applications includes…at least one navigation-related application”, ¶[0006]), which are well known in the art to provide useful information about travel distances and travel time estimates}.
Regarding Claim 4, the combination of Fernandez and Wang discloses all the limitations of the method of Claim 3, as discussed supra. In addition, Fernandez explicitly recites the limitation: wherein the acquiring the secondary response comprises: acquiring weather information of the destination and weather information of the target area {“For the second example input query “Route me to Paris, turn on the wiper, and provide the weather forecast there!”, the first language model 220 may select the applications “routing application”, “vehicle control application”, and “weather application””, ¶[0071]}; and adding the weather information of the destination and the weather information of the target area to the primary response {“The processing unit 240 transmits the second example input query and the applications “routing application”, “vehicle control application”, and “weather application” to the external second language model 243 using the transceiver 242. The external second language model 243 processes the second example input query using the applications “routing application”, “vehicle control application”, and “weather application”. Then, the external second language model 243 transmits a processed second example input query back to the processing unit 240. The processing unit 240 receives the processed second example input query using the transceiver 242 and outputs the processed second example input query to the operation unit 250.”, ¶[0078]}.
Regarding Claim 5, the combination of Fernandez and Wang discloses all the limitations of the method of Claim 2, as discussed supra. In addition, Fernandez explicitly recites the limitation: wherein the acquiring the secondary response comprises: acquiring information on a vehicle type of the vehicle; and adding, based on a place being related to the vehicle type within the destination and the target area, information on the place to the primary response {“evaluating the input query may depend, at least in part, on a location and/or a previous input query and/or a previous conversation and/or a type of a car and/or a specification of a car”, ¶[0042]}; and adding, based on owners of the same vehicle type having visited the destination and the target area, information on a visit frequency to the primary response {“a context used for evaluating the input query may be different from a context used for processing the input query. In particular, a context may comprise one or more of: a location, a previous input query, a previous conversation, a type of a car, a specification of a car, and a battery state of a car.”, ¶[0054]}.
Regarding Claim 9, Fernandez discloses the limitations: an apparatus for controlling operation of a vehicle, the apparatus comprising: a memory configured to store one or more instructions; and at least one processor configured to execute the one or more instructions {computer-implemented method, Abstract} to: acquire, based on a first machine learning model {220, Fig. 2} associated with a current input and a previous stream of inputs {“two example input queries are considered in the following. A first example input query is “Turn on the wiper!”. A second example input query is “Route me to Paris, turn on the wiper, and provide the weather forecast there!”.”, ¶[0067]; also “a previous input query, a previous conversation”, ¶[0054]}, a primary response to the current input {output from first language model 220 into second language model 241, Fig. 2}; acquire, based on applications of a second machine learning model {241, Fig. 2} and a third machine learning model {additional language model in the form of an external language model 243, Fig. 2; and “The plurality of second language models comprises a local second language model 241 and an external second language model 243.”, ¶[0073]} to the primary response, a secondary response {output from language model associated with processing unit 240, Fig. 2}, wherein the second machine learning model is tuned to provision of position information and weather information associated with the vehicle {the second input port to processing unit 240, Fig. 2, can provide additional parametric data to language model 241 via transceiver 242, such as the data types described in ¶[0054] and ¶[0069]}, and wherein the third machine learning model {243, Fig. 2} is tuned to provision of vehicle information {“The plurality of applications may comprise a routing application, a location search application, an electrical vehicle charger search application, a traffic application, a vehicle control application, a weather application, a web search application, a general knowledge application, a feedback application, a car manual application”, ¶[0069]}; output the adjusted secondary response, wherein the current input or the previous stream of inputs is related to a request for information of a destination area or a target area for the vehicle {“example input query is “Route me to Paris, turn on the wiper, and provide the weather forecast there!”, ¶[0067]}; and control, based on the adjusted secondary response, operation of the vehicle {input query 205 begins the process leading to an output from second language model (within processing unit 240) to car control unit 251, Fig. 2}.
Fernandez does not appear to explicitly disclose: adjust, based on a fourth machine learning model, the secondary response, wherein the fourth machine learning model is tuned to a length adjustment of the secondary response or tuned to verification of information associated with the secondary response.
However, Wang explicitly recites the limitation: adjust, based on a fourth machine learning model the secondary response, wherein the fourth machine learning model is tuned to a length adjustment of the secondary response or tuned to verification of information associated with the secondary response {the use of multiple language models combined with fine-tuning, providing the adjusting: “Configuring the language model neural networks and performing a machine learning task can include leveraging the ability of a first large language model to follow prompt-engineered instructions and perform chain-of-thought reasoning, while also fine-tuning a second, smaller language model neural network to optimize the machine learning task performance”, Abstract; see also Fig. 1, which a second prompt and the output of the first language model are combined for inputting into a second language model, ¶[0011]}.
Regarding Claim 10, the combination of Fernandez and Wang discloses all the limitations of the method of Claim 9, as discussed supra. In addition, Fernandez explicitly recites the limitation: wherein the at least one processor is further configured to execute the one or more instructions to: input {input query 205, Fig. 2} a first input for semantic inference to the first machine learning model {“a context used for evaluating the input query may be different from a context used for processing the input query. In particular, a context may comprise one or more of: a location, a previous input query, a previous conversation, a type of a car, a specification of a car, and a battery state of a car.”, ¶[0054]};
input a second input for function classification to the first machine learning model {“In a preferred embodiment, the first language model may perform a classification of the input query.”, ¶[0036]}; input a third input for a query creation to the first machine learning model {“evaluating the input query may depend, at least in part, on a location and/or a previous input query and/or a previous conversation and/or a type of a car and/or a specification of a car”, ¶[0042]}; search, based on the query generated by the first machine learning model, for a document in a database {evaluation of input query requires identification of the vehicle applications, Fig. 3, involved in addressing the input query via apparatus 200 in Fig. 2, which includes “a web search application” (¶[0042]) and other applications requiring searching any onboard databases, look-up tables, etc.}; and input, based on content of the document, a fourth input for generation of the primary response {output from first language model 220, Fig. 2} to the first machine learning model {inputting multiple pieces of data for evaluation by a machine learning model, such a neural network, is well known in the art; additionally, the apparatus of Fernandez in Fig. 2 provides the appropriate output for all the vehicle applications in Fig. 3, which includes vehicle control signals, signals for operating vehicle systems (e.g., wipers 252, sunroof 255 and lights 256), and applications requiring audio, textual or graphical outputs (e.g., infotainment system 261, general knowledge 263 and web search 269.}.
Regarding Claim 11, the combination of Fernandez and Wang discloses all the limitations of the method of Claim 10, as discussed supra. In addition, Fernandez explicitly recites the limitation: wherein the at least one processor is further configured to execute the one or more instructions to: acquire, based on positions of the destination and the target area, an estimated travel time from the destination to the target area; and add the estimated travel time to the primary response {the vehicle applications includes a vehicle navigation application (“the input query comprises selecting one or more applications from a plurality of applications, wherein the plurality of applications includes…at least one navigation-related application”, ¶[0006]), which are well known in the art to provide useful information about travel distances and travel time estimates}.
Regarding Claim 12, the combination of Fernandez and Wang discloses all the limitations of the method of Claim 11, as discussed supra. In addition, Fernandez explicitly recites the limitation: wherein the at least one processor is further configured to execute the one or more instructions to: acquire weather information of the destination and weather information of the target area {“For the second example input query “Route me to Paris, turn on the wiper, and provide the weather forecast there!”, the first language model 220 may select the applications “routing application”, “vehicle control application”, and “weather application””, ¶[0071]}; and add the weather information of the destination and the weather information of the target area to the primary response {“The processing unit 240 transmits the second example input query and the applications “routing application”, “vehicle control application”, and “weather application” to the external second language model 243 using the transceiver 242. The external second language model 243 processes the second example input query using the applications “routing application”, “vehicle control application”, and “weather application”. Then, the external second language model 243 transmits a processed second example input query back to the processing unit 240. The processing unit 240 receives the processed second example input query using the transceiver 242 and outputs the processed second example input query to the operation unit 250.”, ¶[0078]}.
Regarding Claim 13, the combination of Fernandez and Wang discloses all the limitations of the method of Claim 10, as discussed supra. In addition, Fernandez explicitly recites the limitation: wherein the at least one processor is further configured to execute the one or more instructions to: acquire information on a vehicle type of the vehicle; and adding, based on a place being related to the vehicle type within the destination and the target area, information on the place to the primary response {“evaluating the input query may depend, at least in part, on a location and/or a previous input query and/or a previous conversation and/or a type of a car and/or a specification of a car”, ¶[0042]}; and add, based on owners of the same vehicle type having visited the destination and the target area, information on a visit frequency to the primary response {“a context used for evaluating the input query may be different from a context used for processing the input query. In particular, a context may comprise one or more of: a location, a previous input query, a previous conversation, a type of a car, a specification of a car, and a battery state of a car.”, ¶[0054]}.
Claims 6-7 and 14-15 are rejected under 35 U.S.C. §103 as being unpatentable over the combination of Fernandez, Wang and Vasylyev (US 2024/0412720 A1).
Regarding Claim 6, the combination of Fernandez and Wang discloses all the limitations of the method of Claim 1, as discussed supra. The combination of Fernandez and Wang does not appear to explicitly recites the limitation: wherein the adjusting the secondary response comprises: comparing a time remaining until another guidance with a length of the secondary response; and decreasing, based on the length of the secondary response exceeding the time remaining until the other guidance, the length of the secondary response.
However, Vasylyev explicitly recites limitation: wherein the adjusting the secondary response comprises: comparing a time remaining until another guidance with a length of the secondary response; and decreasing, based on the length of the secondary response exceeding the time remaining until the other guidance, the length of the secondary response {taking into account time considerations to adapt to the nature of the conversation between an AI-assistance (Abstract) and a user: “ the processor may be further configured to execute instructions to dynamically adjust the predetermined time period based on at least one of a user input, a system parameter, and a contextual factor.”, ¶[0012], and “Assistant system 2 recalculates the utility score, which now reaches 0.6135, corresponding to a window size of 90 seconds. The system dynamically adjusts the window size between 75 and 90 seconds based on the changing latency and user preferences, balancing the need for context with the real-time demands of the fast-paced conversation.”, ¶[0122]}.
The combination of Fernandez and Wang along with Vasylyev are analogous art because they deal with human-to-machine communication based on artificial intelligence language models.
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of Fernandez and Wang and Vasylyev before them, to modify the teachings of the combination of Fernandez and Wang to include the teachings of Vasylyev to provide a language model based system capable of handling situations requiring a fast response {“Let's now consider a scenario C in which assistant system 2 is balancing user preferences and resource constraints in fast-paced conversations. Imagine the user is engaged in an AI-assisted, fast-paced, interactive conversation with a customer support representative, and this conversation is monitored, analyzed, and augmented on-the-fly by assistant system 2.”, ¶[0121]}.
Regarding Claim 7, the combination of Fernandez and Wang discloses all the limitations of the method of Claim 1, as discussed supra. The combination of Fernandez and Wang does not appear to explicitly recites the limitation: wherein the adjusting the secondary response comprises: increasing or decreasing a length of the secondary response to meet a request from a user of the vehicle, wherein the request is received within the current input and the previous streams of inputs.
However, Vasylyev explicitly recites limitation: wherein the adjusting the secondary response comprises: wherein the adjusting the secondary response comprises: increasing or decreasing a length of the secondary response to meet a request from a user of the vehicle, wherein the request is received within the current input and the previous streams of inputs {taking into account time considerations to adapt to the nature of the conversation between an AI-assistance (Abstract) and a user: “ the processor may be further configured to execute instructions to dynamically adjust the predetermined time period based on at least one of a user input, a system parameter, and a contextual factor.”, ¶[0012], and “balancing the need for context with the real-time demands of the fast-paced conversation.”, ¶[0122]}.
Regarding Claim 14, the combination of Fernandez and Wang discloses all the limitations of the method of Claim 9, as discussed supra. The combination of Fernandez and Wang does not appear to explicitly recites the limitation: wherein the at least one processor is further configured to execute the one or more instructions to: comparing a time remaining until another guidance with a length of the secondary response; and decrease, based on the length of the secondary response exceeding the time remaining until the other guidance, the length of the secondary response.
However, Vasylyev explicitly recites limitation: wherein the at least one processor is further configured to execute the one or more instructions to: comparing a time remaining until another guidance with a length of the secondary response; and decrease, based on the length of the secondary response exceeding the time remaining until the other guidance, the length of the secondary response {taking into account time considerations to adapt to the nature of the conversation between an AI-assistance (Abstract) and a user: “ the processor may be further configured to execute instructions to dynamically adjust the predetermined time period based on at least one of a user input, a system parameter, and a contextual factor.”, ¶[0012], and “Assistant system 2 recalculates the utility score, which now reaches 0.6135, corresponding to a window size of 90 seconds. The system dynamically adjusts the window size between 75 and 90 seconds based on the changing latency and user preferences, balancing the need for context with the real-time demands of the fast-paced conversation.”, ¶[0122]}.
Regarding Claim 15, the combination of Fernandez and Wang discloses all the limitations of the method of Claim 9, as discussed supra. The combination of Fernandez and Wang does not appear to explicitly recites the limitation: wherein the at least one processor is further configured to execute the one or more instructions to: increase or decrease a length of the secondary response to satisfy a request from a user of the vehicle, wherein the request is received within the current input and the previous stream of inputs.
However, Vasylyev explicitly recites limitation: wherein the at least one processor is further configured to execute the one or more instructions to: increase or decrease a length of the secondary response to satisfy a request from a user of the vehicle, wherein the request is received within the current input and the previous stream of inputs {taking into account time considerations to adapt to the nature of the conversation between an AI-assistance (Abstract) and a user: “ the processor may be further configured to execute instructions to dynamically adjust the predetermined time period based on at least one of a user input, a system parameter, and a contextual factor.”, ¶[0012], and “balancing the need for context with the real-time demands of the fast-paced conversation.”, ¶[0122]}.
Claims 8 and 16 are rejected under 35 U.S.C. §103 as being unpatentable over the combination of Fernandez, Wang and Gasser et al. (US 2025/0265347 A1, henceforth Gasser).
Regarding Claim 8, the combination of Fernandez and Wang discloses all the limitations of the method of Claim 1, as discussed supra. The combination of Fernandez and Wang does not appear to explicitly recites the limitation: wherein the adjusting the secondary response comprises: verifying, based on a database used for the primary response and the secondary response, the secondary response; and determining whether a prohibited word is included in the secondary response.
However, Gasser explicitly recites limitation: wherein the adjusting the secondary response comprises: verifying, based on a database used for the primary response and the secondary response {“Data node 104 may store various data, including textual communications, one or more machine learning models (e.g., model weights associated with an LLM, a generative language model, etc.), outputs of machine learning models, semantic data (e.g., textual communications, text files, or embeddings of such text), or training data (e.g., training textual communications or conversations).”, ¶[0021]}, the secondary response; and determining whether a prohibited word is included in the secondary response {As described in ¶[0073], the output of an LLM-based chatbot is monitored to identify and verify prohibited words and prevent their use in the final chatbot output.}.
The combination of Fernandez and Wang along with Gasser are analogous art because they deal with human-to-machine communication based on artificial intelligence language models.
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention, having the teachings of Fernandez and Wang and Gasser before them, to modify the teachings of the combination of Fernandez and Wang to include the teachings of Gasser to identify and prevent prohibited words from being output by a generative language model {¶[0073]}.
Regarding Claim 16, the combination of Fernandez and Wang discloses all the limitations of the method of Claim 9, as discussed supra. The combination of Fernandez and Wang does not appear to explicitly recites the limitation: wherein the at least one processor is further configured to execute the one or more instructions to: verify, based on a database used for the primary response and the secondary response, the secondary response; and determine whether a prohibited word is included in the secondary response
However, Gasser explicitly recites limitation: wherein the at least one processor is further configured to execute the one or more instructions {Fig. 6} to: verify, based on a database used for the primary response and the secondary response {“Data node 104 may store various data, including textual communications, one or more machine learning models (e.g., model weights associated with an LLM, a generative language model, etc.), outputs of machine learning models, semantic data (e.g., textual communications, text files, or embeddings of such text), or training data (e.g., training textual communications or conversations).”, ¶[0021]}, the secondary response; and determine whether a prohibited word is included in the secondary response {As described in ¶[0073], the output of an LLM-based chatbot is monitored to identify and verify prohibited words and prevent their use in the final chatbot output.}.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
US 2025/0328523 A1 – Teaches the use of multiple language models, but distinguishes their approach – termed “prompt tuning” – from fine-tuning in that additional inputs are not added when the output from on language model is fed to the next language model but aim to tune identifying more frequent or important prompts. The model training is less intensive than fine-tuning.
US 2023/0315999 A1- A chatbot approach involving a “fine-tuning method performed in the multi-task deep neural network module” {Fig. 5 and ¶[0025]} with the aim of better identifying the intent of a user’s query.
US 2025/0156760 A1 – Teaches of using multiple language models, in which a core model is supplemented by task-specific models to fine-tune the intermediate outputs form the core model. With this approach, parametric conditions of the core model are unchanged during the iterative process; adjustment so parameters occurs only in the task-specific models.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to RICHARD EDWIN GEIST whose telephone number is (703)756-5854. The examiner can normally be reached Monday-Friday, 9am-6pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Christian Chace can be reached at (571) 272-4190. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/R.E.G./Examiner, Art Unit 3665
/CHRISTIAN CHACE/Supervisory Patent Examiner, Art Unit 3665
Prosecution Insights