Chain-of-draft prompting for generative artificial intelligence models

Abstract

Systems and methods for response generation using a generative artificial intelligence (AI) model are provided. A client device obtains an instructional prompt for a trained generative AI model. The instructional prompt includes a description of a step-by-step reasoning process with a maximum word count for each step. The client device receives a user query. The client device generates a response to the user query using the trained generative AI model based on the instructional prompt.

Description

Atty Docket 1544246—ZM-0880-WOCHAIN-OF-DRAFT PROMPTING FOR GENERATIVE ARTIFICIAL INTELLIGENCE MODELSFIELD

[0001] The present application generally relates to artificial intelligence (Al) and more specifically relates to chain-of- draft (CoD) prompting for generative Al models.BRIEF DESCRIPTION OF THE DRAWINGS

[0002] The accompanying drawings, which are incorporated into and constitute a part of this specification, illustrate one or more certain examples and, together with the description of the example, serve to explain the principles and implementations of the certain examples.

[0003] Figure 1 shows an example system that provides chat and videoconferencing functionalities to various client devices.

[0004] Figure 2 shows an example system in which a chat and video conference provider provides chat and videoconferencing functionalities to various client devices.

[0005] Figure 3 shows an example system that can establish a virtual communication session.

[0006] Figure 4 shows an example system that is configured for response generation and evaluation.

[0007] Figure 5 shows an example GUI displaying a consent authorization request for accessing personal data.

[0008] Figure 6 shows example responses to the same question generated by the same LLM with three different types of instructional prompts.

[0009] Figure 7 shows an example method for generating a response to a user query using a generative Al model with CoD prompting.

[0010] Figure 8 shows an example computing device suitable for use in example systems or methods of generating a response to a user query using a generative Al model according to this disclosure.DETAILED DESCRIPTION

[0011] Examples are described herein in the context of chain-of-draftAtty Docket 1544246--ZM-0880-WO(CoD) prompting for generative artificial intelligence (Al) models. Those of ordinary skill in the art will realize that the following description is illustrative only and is not intended to be in any way limiting. Reference will now be made in detail to implementations of examples as illustrated in the accompanying drawings. The same reference indicators will be used throughout the drawings and the following description to refer to the same or like items.

[0012] In the interest of clarity, not all of the routine features of the examples described herein are shown and described. It will, of course, be appreciated that in the development of any such actual implementation, numerous implementation- specific decisions must be made in order to achieve the developer’s specific goals, such as compliance with application- and business-related constraints, and that these specific goals will vary from one implementation to another and from one developer to another.

[0013] Generative Al models, for example Large Language Models (LLMs), have achieved significant success in solving complex reasoning tasks, often leveraging techniques like Chain of Thought (CoT) prompting. CoT prompting encourages models to break down problems into step-by-step explanations, mimicking the structured reasoning process of humans. While effective, this approach often results in verbose outputs, requiring substantial computational resources and increasing latency. Such verbosity contrasts sharply with how humans typically approach problem- solving. Humans generally rely on concise drafts or shorthand notes to capture essential ideas without unnecessary elaboration during problem solving. For example, when performing complex tasks, such as solving mathematical problems, drafting essays, or coding, a person often jots down only the critical pieces of information that helps the brain progress.

[0014] To reduce computational cost and latency, it is desirable to prompt a generative Al model to draft concise intermediate steps that capture only essential information in the reasoning process. Such a prompt can be called a chain of draft (CoD) prompt.

[0015] For example, a CoD instructional prompt includes an instruction of step-by-step reasoning and a maximum word count (e.g., 5) for each reasoning step. The CoD prompt can also include one or more example queries and corresponding example responses. An example response to an example query includes a chain of one or more example intermediate steps each within theAtty Docket 1544246--ZM-0880-WOmaximum word count and an example final response. An example query is “Jim had 20 apples. He gave Dan some apples. Now Jim has 12 apples. How many apples did Jim give to Dan?’’ The example response is “20 - x = 12; x = 20 - 12. #### 8.’’ The example response concisely represents the reasoning process, without using verbose descriptions as in conventional CoT reasoning.

[0016] With one or more such example query-response pairs in the overall instructional prompt, a generative Al model can generate responses with concise intermediate reasoning steps to user queries, following the format of the example responses. Experiments across a variety of tasks that require multi-step reasoning are conducted to evaluate the effectiveness of CoD prompting. The evaluation demonstrates that this CoD prompting maintains or even improves accuracy in response generation while significantly reducing token usage and latency.

[0017] Thus, the CoD prompting strategy aligns more closely with human reasoning by prioritizing efficiency and minimalism. Instead of verbose intermediate steps, CoD encourages LLMs to generate concise, condensed, yet informative intermediate outputs at each step, so that the generative Al models can focus on advancing toward solutions without the overhead of verbose reasoning. This approach reduces latency and computational costs without sacrificing accuracy. It aligns a generative Al model more closely with humanlike though processes, enhancing their utility in real-time applications with low latency requirements. Overall, CoD prompting has comparable or even improved performance compared to CoT prompting, making generative Al models more practical for real-world applications where efficiency is critical.

[0018] This illustrative example is given to introduce the reader to the general subject matter discussed herein and the disclosure is not limited to this example. The following sections describe various additional non-limiting examples of CoD prompting for generative Al models.

[0019] Referring now to Figure 1, Figure 1 shows an example system 100 that provides chat and videoconferencing functionalities to various client devices. The system 100 includes a chat and video conference provider 110 that is connected to multiple communication networks 120, 130, through which various client devices 140-180 can participate in video conferences hosted by the chat and video conference provider 110. For example, the chat and video conference provider 110 can be located within a private network to provide videoAtty Docket 1544246--ZM-0880-WOconferencing services to devices within the private network, or it can be connected to a public network, e.g., the internet, so it may be accessed by anyone. Some examples may even provide a hybrid model in which a chat and video conference provider 110 may supply components to enable a private organization to host private internal video conferences or to connect its system to the chat and video conference provider 110 over a public network.

[0020] The system optionally also includes one or more authentication and authorization providers, e.g., authentication and authorization provider 115, which can provide authentication and authorization services to users of the client devices 140-160. Authentication and authorization provider 115 may authenticate users to the chat and video conference provider 110 and manage user authorization for the various services provided by chat and video conference provider 110. In this example, the authentication and authorization provider 115 is operated by a different entity than the chat and video conference provider 110, though in some examples, they may be the same entity.

[0021] Chat and video conference provider 110 allows clients to create videoconference meetings (or "meetings") and invite others to participate in those meetings as well as perform other related functionality, such as recording the meetings, generating transcripts from meeting audio, generating summaries and translations from meeting audio, manage user functionality in the meetings, enable text messaging during the meetings, create and manage breakout rooms from the virtual meeting, etc. Figure 2, described below, provides a more detailed description of the architecture and functionality of the chat and video conference provider 110. It should be understood that the term "meeting" encompasses the term "webinar" used herein.

[0022] Meetings in this example chat and video conference provider 110 are provided in virtual rooms to which participants are connected. The room in this context is a construct provided by a server that provides a common point at which the various video and audio data is received before being multiplexed and provided to the various participants. While a "room" is the label for this concept in this disclosure, any suitable functionality that enables multiple participants to participate in a common videoconference may be used.

[0023] To create a meeting with the chat and video conference provider 110, a user may contact the chat and video conference provider 110 using a client device 140-180 and select an option to create a new meeting. Such an option mayAtty Docket 1544246--ZM-0880-WObe provided in a webpage accessed by a client device 140-160 or a client application executed by a client device 140-160. For telephony devices, the user may be presented with an audio menu that they may navigate by pressing numeric buttons on their telephony device. To create the meeting, the chat and video conference provider 110 may prompt the user for certain information, such as a date, time, and duration for the meeting, a number of participants, a type of encryption to use, whether the meeting is confidential or open to the public, etc. After receiving the various meeting settings, the chat and video conference provider may create a record for the meeting and generate a meeting identifier and, in some examples, a corresponding meeting password or passcode (or other authentication information), all of which meeting information is provided to the meeting host.

[0024] After receiving the meeting information, the user may distribute the meeting information to one or more users to invite them to the meeting. To begin the meeting at the scheduled time (or immediately, if the meeting was set for an immediate start), the host provides the meeting identifier and, if applicable, corresponding authentication information (e.g., a password or passcode). The video conference system then initiates the meeting and may admit users to the meeting. Depending on the options set for the meeting, the users may be admitted immediately upon providing the appropriate meeting identifier (and authentication information, as appropriate), even if the host has not yet arrived, or the users may be presented with information indicating that the meeting has not yet started, or the host may be required to specifically admit one or more of the users.

[0025] During the meeting, the participants may employ their client devices 140-180 to capture audio or video information and stream that information to the chat and video conference provider 110. They also receive audio or video information from the chat and video conference provider 110, which is displayed by the respective client device 140 to enable the various users to participate in the meeting.

[0026] At the end of the meeting, the host may select an option to terminate the meeting, or it may terminate automatically at a scheduled end time or after a predetermined duration. When the meeting terminates, the various participants are disconnected from the meeting, and they will no longer receive audio or video streams for the meeting (and will stop transmitting audioAtty Docket 1544246--ZM-0880-WOor video streams). The chat and video conference provider 110 may also invalidate the meeting information, such as the meeting identifier or password / passcode.

[0027] To provide such functionality, one or more client devices 140-180 may communicate with the chat and video conference provider 110 using one or more communication networks, such as network 120 or the public switched telephone network (" PSTN") 130. The client devices 140-180 may be any suitable computing or communication devices that have audio or video capability. For example, client devices 140-160 may be conventional computing devices, such as desktop or laptop computers having processors and computer-readable media, connected to the chat and video conference provider 110 using the internet or other suitable computer network. Suitable networks include the internet, any local area network (" LAN"), metro area network (" MAN"), wide area network (" WAN"), cellular network (e.g., 3G, 4G, 4G LTE, 5G, etc.), or any combination of these. Other types of computing devices may be used instead or as well, such as tablets, smartphones, and dedicated video conferencing equipment. Each of these devices may provide both audio and video capabilities and may enable one or more users to participate in a video conference meeting hosted by the chat and video conference provider 110.

[0028] In addition to the computing devices discussed above, client devices 140-180 may also include one or more telephony devices, such as cellular telephones (e.g., cellular telephone 170), internet protocol (" IP") phones (e.g., telephone 180), or conventional telephones. Such telephony devices may allow a user to make conventional telephone calls to other telephony devices using the PSTN, including the chat and video conference provider 110. It should be appreciated that certain computing devices may also provide telephony functionality and may operate as telephony devices. For example, smartphones typically provide cellular telephone capabilities and thus may operate as telephony devices in the example system 100 shown in Figure 1. In addition, conventional computing devices may execute software to enable telephony functionality, which may allow the user to make and receive phone calls, e.g., using a headset and microphone. Such software may communicate with a PSTN gateway to route the call from a computer network to the PSTN. Thus, telephony devices encompass any devices that can make conventional telephone calls and are not limited solely to dedicated telephony devices like conventional telephones.Atty Docket 1544246--ZM-0880-WO

[0029] Referring again to client devices 140-160, these devices 140-160 contact the chat and video conference provider 110 using network 120 and may provide information to the chat and video conference provider 110 to access functionality provided by the chat and video conference provider 110, such as access to create new meetings or join existing meetings. To do so, the client devices 140-160 may provide user authentication information, meeting identifiers, meeting passwords or passcodes, etc. In examples that employ an authentication and authorization provider 115, a client device, e.g., client devices 140-160, may operate in conjunction with an authentication and authorization provider 115 to provide authentication and authorization information or other user information to the chat and video conference provider 110.

[0030] An authentication and authorization provider 115 may be any entity trusted by the chat and video conference provider 110 that can help authenticate a user to the chat and video conference provider 110 and authorize the user to access the services provided by the chat and video conference provider 110. For example, a trusted entity may be a server operated by a business or other organization with whom the user has created an account, including authentication and authorization information, such as an employer or trusted third-party. The user may sign into the authentication and authorization provider 115, such as by providing a username and password, to access their account information at the authentication and authorization provider 115. The account information includes information established and maintained at the authentication and authorization provider 115 that can be used to authenticate and facilitate authorization for a particular user, irrespective of the client device they may be using. An example of account information may be an email account established at the authentication and authorization provider 115 by the user and secured by a password or additional security features, such as single sign-on, hardware tokens, two-factor authentication, etc. However, such account information may be distinct from functionality such as email. For example, a health care provider may establish accounts for its patients. And while the related account information may have associated email accounts, the account information is distinct from those email accounts.

[0031] Thus, a user's account information relates to a secure, verified set of information that can be used to authenticate and provide authorization services for a particular user and should be accessible only by that user. ByAtty Docket 1544246--ZM-0880-WOproperly authenticating, the associated user may then verify themselves to other computing devices or services, such as the chat and video conference provider 110. The authentication and authorization provider 115 may require the explicit consent of the user before allowing the chat and video conference provider 110 to access the user's account information for authentication and authorization purposes.

[0032] Once the user is authenticated, the authentication and authorization provider 115 may provide the chat and video conference provider 110 with information about services the user is authorized to access. For instance, the authentication and authorization provider 115 may store information about user roles associated with the user. The user roles may include collections of services provided by the chat and video conference provider 110 that users assigned to those user roles are authorized to use. Alternatively, more or less granular approaches to user authorization may be used.

[0033] When the user accesses the chat and video conference provider 110 using a client device, the chat and video conference provider 110 communicates with the authentication and authorization provider 115 using information provided by the user to verify the user's account information. For example, the user may provide a username or cryptographic signature associated with an authentication and authorization provider 115. The authentication and authorization provider 115 then either confirms the information presented by the user or denies the request. Based on this response, the chat and video conference provider 110 either provides or denies access to its services, respectively.

[0034] For telephony devices, e.g., client devices 170-180, the user may place a telephone call to the chat and video conference provider 110 to access video conference services. After the call is answered, the user may provide information regarding a video conference meeting, e.g., a meeting identifier (" ID"), a passcode or password, etc., to allow the telephony device to join the meeting and participate using audio devices of the telephony device, e.g., microphone(s) and speaker(s), even if video capabilities are not provided by the telephony device.

[0035] Because telephony devices typically have more limited functionality than conventional computing devices, they may be unable to provide certain information to the chat and video conference provider 110. For example, telephony devices may be unable to provide authentication informationAtty Docket 1544246--ZM-0880-WOto authenticate the telephony device or the user to the chat and video conference provider 110. Thus, the chat and video conference provider 110 may provide more limited functionality to such telephony devices. For example, the user may be permitted to join a meeting after providing meeting information, e.g.. a meeting identifier and passcode, but only as an anonymous participant in the meeting. This may restrict their ability to interact with the meetings in some examples, such as by limiting their ability to speak in the meeting, hear or view certain content shared during the meeting, or access other meeting functionality, such as joining breakout rooms or engaging in text chat with other participants in the meeting.

[0036] It should be appreciated that users may choose to participate in meetings anonymously and decline to provide account information to the chat and video conference provider 110, even in cases where the user could authenticate and employs a client device capable of authenticating the user to the chat and video conference provider 110. The chat and video conference provider 110 may determine whether to allow such anonymous users to use services provided by the chat and video conference provider 110. Anonymous users, regardless of the reason for anonymity, may be restricted as discussed above with respect to users employing telephony devices, and in some cases may be prevented from accessing certain meetings or other services, or may be entirely prevented from accessing the chat and video conference provider 110.

[0037] Referring again to chat and video conference provider 110, in some examples, it may allow client devices 140-160 to encrypt their respective video and audio streams to help improve privacy in their meetings. Encryption may be provided between the client devices 140-160 and the chat and video conference provider 110 or it may be provided in an end-to-end configuration where multimedia streams (e.g., audio or video streams) transmitted by the client devices 140-160 are not decrypted until they are received by another client device 140-160 participating in the meeting. Encryption may also be provided during only a portion of a communication, for example encryption may be used for otherwise unencrypted communications that cross international borders.

[0038] Client- to- server encryption may be used to secure the communications between the client devices 140-160 and the chat and video conference provider 110, while allowing the chat and video conference provider 110 to access the decrypted multimedia streams to perform certain processing,Atty Docket 1544246--ZM-0880-WOsuch as recording the meeting for the participants or generating transcripts of the meeting for the participants. End-to-end encryption may be used to keep the meeting entirely private to the participants without any worry about a chat and video conference provider 110 having access to the substance of the meeting. Any suitable encryption methodology may be employed, including key-pair encryption of the streams. For example, to provide end-to-end encryption, the meeting host's client device may obtain public keys for each of the other client devices participating in the meeting and securely exchange a set of keys to encrypt and decrypt multimedia content transmitted during the meeting. Thus, the client devices 140-160 may securely communicate with each other during the meeting. Further, in some examples, certain types of encryption may be limited by the types of devices participating in the meeting. For example, telephony devices may lack the ability to encrypt and decrypt multimedia streams. Thus, while encrypting the multimedia streams may be desirable in many instances, it is not required as it may prevent some users from participating in a meeting.

[0039] By using the example system shown in Figure 1, users can create and participate in meetings using their respective client devices 140-180 via the chat and video conference provider 110. Further, such a system enables users to use a wide variety of different client devices 140-180 from traditional standards-based video conferencing hardware to dedicated video conferencing equipment to laptop or desktop computers to handheld devices to legacy telephony devices, etc.

[0040] Referring now to Figure 2, Figure 2 shows an example system 200 in which a chat and video conference provider 210 provides chat and videoconferencing functionalities to various client devices 220-250. The client devices 220-250 include two conventional computing devices 220-230. dedicated equipment for a video conference room 240, and a telephony device 250. Each client device 220-250 communicates with the chat and video conference provider 210 over a communications network, such as the internet for client devices 220-240 or the PSTN for client device 250, generally as described above with respect to Figure 1. The chat and video conference provider 210 is also in communication with one or more authentication and authorization providers 215, which can authenticate various users to the chat and video conference provider 210 generally as described above with respect to Figure 1.

[0041] In this example, the chat and video conference provider 210 employs multiple different servers (or groups of servers) to provide differentAtty Docket 1544246-ZM-0880-WOexamples of video conference functionality, thereby enabling the various client devices to create and participate in video conference meetings. The chat and video conference provider 210 uses one or more real-time media servers 212, one or more network services servers 214, one or more video room gateways 216. one or more message and presence gateways 217, and one or more telephony gateways 218. Each of these servers 212-218 is connected to one or more communications networks to enable them to collectively provide access to and participation in one or more video conference meetings to the client devices 220-250.

[0042] The real-time media servers 212 provide multiplexed multimedia streams to meeting participants, such as the client devices 220-250 shown in Figure 2. While video and audio streams typically originate at the respective client devices, they are transmitted from the client devices 220-250 to the chat and video conference provider 210 via one or more networks where they are received by the real-time media servers 212. The real-time media servers 212 determine which protocol is optimal based on, for example, proxy settings and the presence of firewalls, etc. For example, the client device might select among UDP. TCP, TLS, or HTTPS for audio and video and UDP for content screen sharing.

[0043] The real-time media servers 212 then multiplex the various video and audio streams based on the target client device and communicate multiplexed streams to each client device. For example, the real-time media servers 212 receive audio and video streams from client devices 220-240 and only an audio stream from client device 250. The real-time media servers 212 then multiplex the streams received from devices 230-250 and provide the multiplexed stream to client device 220. The real-time media servers 212 are adaptive, for example, reacting to real-time network and client changes, in how they provide these streams. For example, the real-time media servers 212 may monitor parameters such as a client's bandwidth CPU usage, memory and network I / O as well as network parameters such as packet loss, latency and jitter to determine how to modify the way in which streams are provided.

[0044] The client device 220 receives the stream, performs any decryption, decoding, and demultiplexing on the received streams, and then outputs the audio and video using the client device's video and audio devices. In this example, the real-time media servers do not multiplex client device 220's ownAtty Docket 1544246-ZM-0880-WOvideo and audio feeds when transmitting streams to it. Instead, each client device 220-250 only receives multimedia streams from other client devices 220-250. For telephony devices that lack video capabilities, e.g., client device 250, the realtime media servers 212 only deliver multiplex audio streams. The client device 220 may receive multiple streams for a particular communication, allowing the client device 220 to switch between streams to provide a higher quality of service.

[0045] In addition to multiplexing multimedia streams, the real-time media servers 212 may also decrypt incoming multimedia stream in some examples. As discussed above, multimedia streams may be encrypted between the client devices 220-250 and the chat and video conference provider 210. In some such examples, the real-time media servers 212 may decrypt incoming multimedia streams, multiplex the multimedia streams appropriately for the various clients, and encrypt the multiplexed streams for transmission.

[0046] As mentioned above with respect to Figure 1, the chat and video conference provider 210 may provide certain functionality with respect to unencrypted multimedia streams at a user's request. For example, the meeting host may be able to request that the meeting be recorded or that a transcript of the audio streams be prepared, which may then be performed by the real-time media servers 212 using the decrypted multimedia streams, or the recording or transcription functionality may be off-loaded to a dedicated server (or servers), e.g.. cloud recording servers, for recording the audio and video streams. In some examples, the chat and video conference provider 210 may allow a meeting participant to notify it of inappropriate behavior or content in a meeting. Such a notification may trigger the real-time media servers to 212 record a portion of the meeting for review by the chat and video conference provider 210. Still other functionality may be implemented to take actions based on the decrypted multimedia streams at the chat and video conference provider, such as monitoring video or audio quality, adjusting or changing media encoding mechanisms, etc.

[0047] It should be appreciated that multiple real-time media servers 212 may be involved in communicating data for a single meeting and multimedia streams may be routed through multiple different real-time media servers 212. In addition, the various real-time media servers 212 may not be co-located. but instead may be located at multiple different geographic locations, which may enable high-quality communications between clients that are dispersed over wideAtty Docket 1544246-ZM-0880-WOgeographic areas, such as being located in different countries or on different continents. Further, in some examples, one or more of these servers may be colocated on a client's premises, e.g., at a business or other organization. For example, different geographic regions may each have one or more real-time media servers 212 to enable client devices in the same geographic region to have a high-quality connection into the chat and video conference provider 210 via local servers 212 to send and receive multimedia streams, rather than connecting to a real-time media server located in a different country or on a different continent. The local real-time media servers 212 may then communicate with physically distant servers using high-speed network infrastructure, e.g., internet backbone network(s), that otherwise might not be directly available to client devices 220-250 themselves. Thus, routing multimedia streams may be distributed throughout the video conference system and across many different real-time media servers 212.

[0048] Turning to the network services servers 214, these servers 214 provide administrative functionality to enable client devices to create or participate in meetings, send meeting invitations, create or manage user accounts or subscriptions, and other related functionality. Further, these servers may be configured to perform different functionalities or to operate at different levels of a hierarchy, e.g., for specific regions or localities, to manage portions of the chat and video conference provider under a supervisory set of servers. When a client device 220-250 accesses the chat and video conference provider 210, it will typically communicate with one or more network services servers 214 to access their account or to participate in a meeting.

[0049] When a client device 220-250 first contacts the chat and video conference provider 210 in this example, it is routed to a network services server 214. The client device may then provide access credentials for a user, e.g., a username and password or single sign-on credentials, to gain authenticated access to the chat and video conference provider 210. This process may involve the network services servers 214 contacting an authentication and authorization provider 215 to verify the provided credentials. Once the user's credentials have been accepted, and the user has consented, the network services servers 214 may perform administrative functionality, like updating user account information, if the user has account information stored with the chat and video conference provider 210. or scheduling a new meeting, by interacting with the networkAtty Docket 1544246-ZM-0880-WOservices servers 214. Authentication and authorization provider 215 may be used to determine which administrative functionality a given user may access according to assigned roles, permissions, groups, etc.

[0050] In some examples, users may access the chat and video conference provider 210 anonymously. When communicating anonymously, a client device 220-250 may communicate with one or more network services servers 214 but only provide information to create or join a meeting, depending on what features the chat and video conference provider allows for anonymous users. For example, an anonymous user may access the chat and video conference provider using client device 220 and provide a meeting ID and passcode. The network services server 214 may use the meeting ID to identify an upcoming or on-going meeting and verify the passcode is correct for the meeting ID. After doing so, the network services server(s) 14 may then communicate information to the client device 220 to enable the client device 220 to join the meeting and communicate with appropriate real-time media servers 212.

[0051] In cases where a user wishes to schedule a meeting, the user (anonymous or authenticated) may select an option to schedule a new meeting and may then select various meeting options, such as the date and time for the meeting, the duration for the meeting, a type of encryption to be used, one or more users to invite, privacy controls (e.g.. not allowing anonymous users, preventing screen sharing, manually authorize admission to the meeting, etc.), meeting recording options, etc. The network services servers 214 may then create and store a meeting record for the scheduled meeting. When the scheduled meeting time arrives (or within a threshold period of time in advance), the network services server(s) 14 may accept requests to join the meeting from various users.

[0052] To handle requests to join a meeting, the network services server(s) 214 may receive meeting information, such as a meeting ID and passcode, from one or more client devices 220-250. The network services server(s) 214 locate a meeting record corresponding to the provided meeting ID and then confirm whether the scheduled start time for the meeting has arrived, whether the meeting host has started the meeting, and whether the passcode matches the passcode in the meeting record. If the request is made by the host, the network services server(s) 214 activates the meeting and connects the host to a real-time media server 212 to enable the host to begin sending and receiving multimediaAtty Docket 1544246-ZM-0880-WOstreams.

[0053] Once the host has started the meeting, subsequent users requesting access will be admitted to the meeting if the meeting record is located and the passcode matches the passcode supplied by the requesting client device 220-250. In some examples additional access controls may be used as well. But if the network services server(s) 214 determines to admit the requesting client device 220-250 to the meeting, the network services server 214 identifies a realtime media server 212 to handle multimedia streams to and from the requesting client device 220-250 and provides information to the client device 220-250 to connect to the identified real-time media server 212. Additional client devices 220-250 may be added to the meeting as they request access through the network services server(s) 214.

[0054] After joining a meeting, client devices will send and receive multimedia streams via the real-time media servers 212. but they may also communicate with the network services servers 214 as needed during meetings. For example, if the meeting host leaves the meeting, the network services server(s) 214 may appoint another user as the new meeting host and assign host administrative privileges to that user. Hosts may have administrative privileges to allow them to manage their meetings, such as by enabling or disabling screen sharing, muting or removing users from the meeting, assigning or moving users to the mainstage or a breakout room if present, recording meetings, etc. Such functionality may be managed by the network services server(s) 214.

[0055] For example, if a host wishes to remove a user from a meeting, they may select a user to remove and issue a command through a user interface on their client device. The command may be sent to a network services server 214, which may then disconnect the selected user from the corresponding realtime media server 212. If the host wishes to remove one or more participants from a meeting, such a command may also be handled by a network services server 214, which may terminate the authorization of the one or more participants for joining the meeting.

[0056] In addition to creating and administering on-going meetings, the network services server(s) 214 may also be responsible for closing and tearingdown meetings once they have been completed. For example, the meeting host may issue a command to end an on-going meeting, which is sent to a network services server 214. The network services server 214 may then remove anyAtty Docket 1544246-ZM-0880-WOremaining participants from the meeting, communicate with one or more real time media servers 212 to stop streaming audio and video for the meeting, and deactivate, e.g., by deleting a corresponding passcode for the meeting from the meeting record, or delete the meeting record(s) corresponding to the meeting. Thus, if a user later attempts to access the meeting, the network services server(s) 214 may deny the request.

[0057] Depending on the functionality provided by the chat and video conference provider, the network services server(s) 214 may provide additional functionality, such as by providing private meeting capabilities for organizations, special types of meetings (e.g., webinars), etc. Such functionality may be provided according to various examples of video conferencing providers according to this description.

[0058] Referring now to the video room gateway servers 16. these servers 216 provide an interface between dedicated video conferencing hardware, such as may be used in dedicated video conferencing rooms. Such video conferencing hardware may include one or more cameras and microphones and a computing device designed to receive video and audio streams from each of the cameras and microphones and connect with the chat and video conference provider 210. For example, the video conferencing hardware may be provided by the chat and video conference provider to one or more of its subscribers, which may provide access credentials to the video conferencing hardware to use to connect to the chat and video conference provider 210.

[0059] The video room gateway servers 216 provide specialized authentication and communication with the dedicated video conferencing hardware that may not be available to other client devices 220-230, 250. For example, the video conferencing hardware may register with the chat and video conference provider when it is first installed and the video room gateway may authenticate the video conferencing hardware using such registration as well as information provided to the video room gateway server(s) 216 when dedicated video conferencing hardware connects to it, such as device ID information, subscriber information, hardware capabilities, hardware version information etc. Upon receiving such information and authenticating the dedicated video conferencing hardware, the video room gateway server(s) 216 may interact with the network services servers 214 and real-time media servers 212 to allow the video conferencing hardware to create or join meetings hosted by the chat andAtty Docket 1544246-ZM-0880-WOvideo conference provider 210.

[0060] Referring now to the telephony gateway servers 218, these servers 218 enable and facilitate telephony devices' participation in meetings hosted by the chat and video conference provider 10. Because telephony devices communicate using the PSTN and not using computer networking protocols, such as TCP / IP, the telephony gateway servers 218 act as an interface that converts between the PSTN, and the networking system used by the chat and video conference provider 210.

[0061] For example, if a user uses a telephony device to connect to a meeting, they may dial a phone number corresponding to one of the chat and video conference provider's telephony gateway servers 218. The telephony gateway server 218 will answer the call and generate audio messages requesting information from the user, such as a meeting ID and passcode. The user may enter such information using buttons on the telephony device, e.g., by sending dual-tone multi-frequency (" DTMF") audio streams to the telephony gateway server 218. The telephony gateway server 218 determines the numbers or letters entered by the user and provides the meeting ID and passcode information to the network services servers 214, along with a request to join or start the meeting, generally as described above. Once the telephony client device 250 has been accepted into a meeting, the telephony gateway server is instead joined to the meeting on the telephony device's behalf.

[0062] After joining the meeting, the telephony gateway server 218 receives an audio stream from the telephony device and provides it to the corresponding real-time media server 212 and receives audio streams from the real-time media server 212, decodes them, and provides the decoded audio to the telephony device. Thus, the telephony gateway servers 218 operate essentially as client devices, while the telephony device operates largely as an input / output device, e.g.. a microphone and speaker, for the corresponding telephony gateway server 218, thereby enabling the user of the telephony device to participate in the meeting despite not using a computing device or video.

[0063] It should be appreciated that the components of the chat and video conference provider 210 discussed above are merely examples of such devices and an example architecture. Some video conference providers may provide more or less functionality than described above and may not separate functionality into different types of servers as discussed above. Instead, any suitable servers andAtty Docket 1544246-ZM-0880-WOnetwork architectures may be used according to different examples.

[0064] Referring now to Figure 3, Figure 3 shows an example system 300 for using LLMs for meta prompt optimization and task prompt enhancement. In this example, the system 300 includes a client device 330, a communication platform 310. and one or more remote servers 380 that host one or more Al models 382 in network communication with network 320. In this example, the communication platform 310 provides chat and virtual conferencing capabilities, such as discussed above with respect to Figures 1-2, but also provides one or more servers 312 that provide one or more Al models 314 that may be used to service requests received from users via their respective client devices, such as client device 330. In addition, the communication platform 310 provides a response generator 316 to generate a response to a user query.

[0065] The one or more Al models 314 may include a model that has been trained on a large corpus of data, such as information available from licensed, commercially usable, non-public datasets. For LLMs, the training data may be written materials, such as webpages, documents, emails, or blogs that may be relevant to generating written works.

[0066] Client devices 330 may execute client software 332 to join and participate in virtual conferences hosted by the communication platform 310. During a virtual conference, the participants can exchange audio and video streams, as discussed above with respect to Figures 1-2, to interact with each other, discuss any topics of interest, and share content. Similarly, the participants can continue any discussions outside of a virtual conference, such as by using chat functionality provided by the communication platform 310. They may also email each other using email services provided by the communication platform 310 or another third party.

[0067] In some examples, a user of the communication platform 310 wants to obtain some information based on certain communication data using a generative model. For example, a user wants to use a trained ML model, such as an LLM, to perform a generative task, for example, to generate a summary of the meeting transcript. Thus, a user needs to provide a task prompt to the trained AIL model as instructions. However, user provided task prompts may not be written with clear structure and specific information used by LLMs, thus the output may not be as good as when a well written task prompt is provided.Atty Docket 1544246-ZM-0880-WO

[0068] To help with this process, the communication platform 310 can receive the user's task prompt and attempt to enhance it by using an LLM to generate an enhanced version of the user's task prompt. The communication platform 310 can do this by generating multiple meta prompts from the initial meta prompt and evaluating the different meta prompts. The communication platform 310 can then select the best meta prompt to provide to the LLM as a prompt for enhancing the user’s task prompt. The enhanced task prompt can then be provided to an LLM to cause it to perform the user’s requested generative task.

[0069] Referring now to Figure 3, Figure 3 shows an example system 300 for using generative Al models to perform certain tasks on a communication platform. In this example, the system 300 includes a client device 330, a communication platform 310. and one or more remote servers 380 in network communication with network 320. In this example, the communication platform 310 provides chat and virtual conferencing capabilities, such as discussed above with respect to Figures 1-2, but also provides one or more servers 312 that uses one or more Al models 314 to service requests received from users via their respective client device, such as client device 330. The Al models 314 includes a model that has been trained or fine-tuned on a large corpus of data, such as information available from licensed, commercially usable, non-public datasets. In addition, the communication platform 310 provides a response generator 316 to respond to a user query. The one or more remote servers 380 host one or more Al models 382, which can be used by the servers 312 on the communication platform 310, in addition to or as alternatives to the Al models 314, to service requests received from users.

[0070] The Al models 314 or 382 can include various types of AI / ML models. Example Al models 314 includes simple ML models, such as Linear Regression and Gradient Boosting, or more sophisticated models, such as Factorization Machines (“FM”). The Al models 314 also include deep learning models, such as DeepFM and Wide& Deep or other similar models. Other alternative ML models include a deep convolutional neural network (“CNN”), a residual neural network (“ResNet"), or a recurrent neural network (“RNN”), e.g.. long short-term memory (“LSTM") models or gated recurrent units (“GRUs") models. The ML model can also be any other suitable ML model, such as a three-dimensional CNN (“3DCNN”), a dynamic time warping (“DTW”) technique, aAtty Docket 1544246-ZM-0880-WOhidden Markov model (“HMM”), etc., or combinations of one or more of such techniques — e.g., CNN-HMM or MCNN (Multi-Scale Convolutional Neural Network). Further, some Al models 314 include adversarial networks, such as generative adversarial networks (“GANs”), or may employ autoencoders CAEs’ ) in conjunction with ML models, such as AEGANs or variational AEGANs (“VAEGANs”). Some Al models 314 may use transformer networks or selfattention based neural networks. Some Al models 314 are generative Al models, such as generative pre-trained transformer (GPT), Text-to-Text Transfer Transformer (T5), Bidirectional and Auto-Regressive Transformer (BART), Bidirectional Encoder Representations from Transformer (BERT), their variations, or other large language models (LLMs) or foundation models (FM). The Al models 314 can be supervised or unsupervised learning models. LLMs can include regular LLMs. which can have hundreds of billions or even trillions of parameters. The LLMs can also include smaller versions of LLMs. which can have a few million to a few billion parameters. Examples of LLMs include GPT models of different types, autoregressive LLMs (e.g., Large Language Model Meta A (LLaMA)), transformer-based autoregressive LLMs (e.g., BigScience Large Open-science Open-access Multilingual Language Models (BLOOMs)). Zephyr, MISTRAL, causal decoder-only models (e.g., Falcon), MosaicML Pretrained Transformer (MPT) models, Bidirectional Encoder Representations from Transformers (BERT) models, or T5 models. Especially, some examples of small LLMs can include DistilBERT, BERT Mini. BERT Small, BERT Medium, BERT Tiny, MobileBERT, T5-small, Orca 2, GPT- Neo, GPT-J, all of which are scaled-down versions of regular LLMs.

[0071] Client devices 330 may execute client software 332 to join and participate in virtual conferences hosted by the communication platform 310. During a virtual conference, the participants can exchange audio and video streams, as discussed above with respect to Figures 1-2, to interact with each other, discuss any topics of interest, and share content. Similarly, the participants can continue any discussions outside of a virtual conference, such as by using chat functionality provided by the communication platform 310. They may also email each other using email services provided by the communication platform 310 or another third party.

[0072] A user may provide a user query via a client device 330. The user query can be provided to the response generator 316 on the communicationAtty Docket 1544246--ZM-0880-WOplatform 310, which uses one or more Al models from the Al models 314 or 382. The response generator 316 can generate a response to the user query.

[0073] Referring now to Figure 4, Figure 4 shows an example system that is configured for response generation and evaluation. The communication platform 310 is in network communication with a client device 330. The communication platform 310 includes a data store 410. a model store 420, an AI-powered response generator 430, and an evaluation engine 440.

[0074] The data store 410 stores communication data associated with different users or different communication sessions, user profile data, and other suitable types of data. The communication data includes video conference recordings, video conference transcripts, chat messages, emails, and other suitable types of communication data. The user data includes user profile data and activity data. User profile data includes name, job title, location, etc. User activity data includes communication activities associated with a user. The data store 410 also stores user queries and responses generated by the Al-powered response generator 430 with CoD prompting. In some examples, the data store 410 also includes responses generated by the Al-powered response generator 430 using CoT prompting or no prompting. The data store 410 also includes evaluation data generated by the evaluation engine 440. The evaluation data is associated with performance of the Al-powered response generator 430 using different prompting strategies or different Al models.

[0075] The model store 420 stores AI / ML models that can be used by the Al-powered response generator 430. In some examples, the model store 420 stores the Al models on the communication platform 310. In some examples, the model store 420 includes APIs for accessing various Al models from the communication platform 310, remote servers 380, or other third-party platforms.

[0076] The Al -powered response generator 430 is configured to generate responses to user queries using a generative Al model. The generative Al model can be selected from the model store 420. While any suitable Al model can be used, as discussed above in Figure 3, the Al models used by the Al-powered response generator 430 are usually LLMs, for example GPT models or other transformer-based models that have generative capabilities. The generative Al model is configured with an overall instructional prompt. The overall instructional prompt is predetermined by an administrator user as a system prompt for the generative Al model and can be updated by the administratorAtty Docket 1544246--ZM-0880-WOuser. The overall instructional prompt includes an instruction of step-by-step reasoning and a maximum word count (e.g., 5) for each reasoning step. For example, the overall instructional prompt includes “think step by step and only keep a minimum draft for each thinking step with 5 words at most. Return the final answer at the end of the response after a separator ####.” The maximum word count provides as a general guideline not a strict rule for the generative Al model to generate short reasoning steps. Some intermediate steps may go over the maximum word count if necessary.

[0077] In some examples, the overall instructional prompt also includes one or more example user queries and corresponding examples responses. An example query is “Jim had 20 apples. He gave Dan some apples. Now Jim has 12 apples. How many apples did Jim give to Dan?” The example response is “20 - x = 12; x = 20 - 12. #### 8.” Another example query is “A coin is heads up. Ann flips the coin. Blair does not flip the coin. Connie flips the coin. Is the coin still heads up?” The example response is “first, heads up; second, flip, tails up; third, no flip, tails up; fourth, flip, heads up. #### Yes.” Each of the example responses includes a chain of draft intermediate steps within the maximum word count, which concisely represents the reasoning process. An administrator user can add, remove, or edit the example queries and example responses in the overall instructional prompt. In some examples, the generative Al model used in the AI-powered response generator 430 is fine-tuned with training data, including training user queries and training responses, similar to the example user queries and corresponding example responses as described above. The training responses include training intermediate steps within the maximum word count.

[0078] Recent development in structured reasoning methods, such as CoT-based reasoning, Tree of Thought, Graph of Thought, Reasoning + Action (ReAct), greatly enhances Al models' ability to solve complex questions. However, these structured reasoning processes also significantly increase the token usage before arriving at a final answer. This makes it challenging to apply in cost sensitive and latency-sensitive scenarios. Furthermore, the model’s lack of awareness regarding task complexity often leads to overthinking even on simple tasks, resulting in unnecessary resource consumption. Certain existing approaches try to tackle the resource cost and latency drawbacks. However, either they can only address one of the issues, or they add additional issues, such as reduced accuracy or limiting applications.Atty Docket 1544246--ZM-0880-WO

[0079] In contrast, example CoD approaches according to this disclosure may employ a per-step budget, allowing unlimited reasoning steps, which makes it more adaptable to various structured reasoning techniques. Such CoD prompting approaches better mimics human thought process. Humans tend to adopt a more concise approach when solving complex problems involving multi-step reasoning, such as mathematical or logical puzzles. Rather than elaborating on every detail, humans typically jot down only the essential intermediate results, for example minimal drafts, to facilitate their thought processes.

[0080] The evaluation engine 440 is configured to evaluate the performance of a generative Al model with different overall prompts on different types of tasks. In an example, the evaluation engine 440 evaluates reasoning performances on three categories of tasks by different LLMs with different prompting paradigms. The three categories of tasks are associated with arithmetic reasoning, common- sense reasoning, and symbolic reasoning, respectively. These tasks can be from one or more existing data sets, which include questions and reference answers. The evaluation engine 440 deploys two different types of LLMs, for example a GPT-4o model and a Claude 3.5 Sonnet model, to perform each task. Three different types of system prompts are provided to each LLM for performing each task. An example baseline system prompt is “Answer the question directly. Do not return any preamble. explanation, or reasoning.” An example CoT system prompt is “Think step by step to answer the following question. Return the answer at the end of the response after a separator ####.” An example CoD system prompt is “Think step by step but only keep a minimum draft for each thinking step, with 5 words at most. Return the answer at the end of the response after a separator ####.” Each system prompt can also include one or more few-shot examples, such as inputoutput pairs. Alternatively, a user query (a task prompt) can include an example query and example response, where the example query is similar to the user query.

[0081] The evaluation engine 440 determines the accuracy of the responses by comparing the generated responses with corresponding reference answers from the existing data sets. The evaluation engine 440 also counts the number of tokens in each generated response. For example, the evaluation engine 440 uses a tokenizer to tokenize a generated response and uses a counterAtty Docket 1544246--ZM-0880-WOto count the number of tokens. The evaluation engine 440 also measures the time that a generative Al model takes to generate a response to represent latency.

[0082] Performances by different Al models with different system prompts on arithmetic reasoning tasks are presented in Table 1. The arithmetic reasoning tasks pose significant challenges for both the GPT-4o model and the Claude 3.5 Sonnet model when using the standard baseline prompt, yielding accuracies of 53.3% and 64.6%, respectively. However, with the CoT prompt, both models surpass 95% accuracy, albeit at the expense of generating approximately 200 tokens per response. In contrast, the two models with the CoD prompt achieve a comparable accuracy of 91% while requiring only about 40 tokens per response, thereby reducing the average output token count by 80% and cutting the average latency by 76.2% and 48.4%, respectively.Table 1. Evaluation Results for Arithmetic Reasoning Tasks.Model Prompt Accuracy Token # Latency Standard 53.3% 1.1 0.6 sGPT-4o CoT 95.4% 205.1 4.2 sCoD 91.1% 43.9 1.0 s„ Standard 64.6% 1.1 0.9 s Claude. J.. JCoT 95.8% 190.0 3.1 s Sonnet CoD 91.4% 39.8 1.6 s

[0083] Date understanding tasks and sports understanding tasks are selected to evaluate performances of different models with different system prompts on common sense reasoning. Date understanding tasks involve inferring a date from a given context. An example date understanding task is “The first day of 2019 is a Tuesday, and today is the first Monday of 2019. What is the date today in MM / DD / YYYY?” Sports understanding tasks involve determining whether a sentence relating to sports is plausible or implausible. An example sports understanding task is “Is the following sentence plausible? ‘Kyle Palmieri was called for slashing. For consistency, the same system prompts employed in the arithmetic reasoning evaluation above are provided to the same models for common sense reasoning evaluation.

[0084] The evaluation results for LLM performances on date understanding reasoning tasks with different prompts are presented in Table 2. The evaluation results for LLM performances on sports understanding reasoning tasks with different prompts are presented in Table 3. Both Table 2 and Table 3Atty Docket 1544246--ZM-0880-WOshow that CoD significantly reduces both latency and cost by generating considerably fewer tokens in responses compared to CoT. Additionally, CoD outperforms CoT in accuracy in various cases. Notably, CoT prompting leads the Claude 3.5 Sonnet model to generate excessively verbose responses, especially in the sports understanding tasks, where CoD reduces the average output tokens from 189.4 to 14.3. which is a 92.4% reduction.Table 2. Evaluation Results for Date Understanding Tasks. Model Prompt Accuracy Token # Latency Standard 72.6% 5.2 0.6 sGPT-4o CoT 90.2% 75.7 1.7 sCoD 88.1% 30.2 1.3 s_ Standard 84.3% 5.2 1.0 s Claude.,..)CoT 87.0% 172.5 3.2 s Sonnet CoD 89.7% 31.3 1.4 sTable 3. Evaluation Results for Sports Understanding Tasks.Model Prompt Accuracy Token # Latency Standard 90.0% 1.0 0.4 sGPT-4o CoT 95.9% 28.7 0.9 sCoD 98.3% 15.0 0.7 s„ 1 c - Standard 90.6% 1.0 0.9 s Claude 0.0 CoT 93.2% 189.4 3.6 s Sonnet CoD 97.3% 14.3 1.0 s

[0085] Symbolic reasoning tasks include coin flipping related tasks. In an example coin flipping task, the LLMs are asked to predict which side is up after a sequence of coin flip actions. An example coin flipping task is ““A coin is heads up. Ann flips the coin. Blair does not flip the coin. Connie flips the coin. Is the coin still heads up?” The evaluation results for GPT-4o and Claude 3.5 Sonnet are shown in Table 4. They achieve 73.2% and 85.2% with standard prompting, respectively. However, both models reach a perfect 100% accuracy with CoT and CoD. Again, CoD demonstrates significant reduction of tokens compared to CoT, from 68% for GPT-4o to 86% for Claude 3.5 Sonnet.Table 4. Evaluation Results for Sports Understanding Tasks.Model Prompt Accuracy Token # Latency Standard 73.2% 1.0 0.4 sGPT-4o CoT 100.0% 52.4 1.4 sCoD 100.0% 16.8 0.8 s Claude 3.5 Standard 85.2% 1.0 1.2 s Sonnet CoT 100.0% 135.3 3.1 sAtty Docket 1544246-ZM-0880-WOCoD 100.0% 18.9 1.6 s

[0086] The evaluation results by the evaluation engine 440 demonstrate that the present CoD prompting techniques substantially reduce the latency required for reasoning while achieving comparable — and in some cases even superior — accuracy compared to traditional CoD prompting strategies. Unlike traditional CoT methods that often involve lengthy reasoning steps. CoD prompting leverages concise reasoning drafts to speed up response generation without sacrificing correctness.

[0087] Additionally, reasoning based on CoD prompts offers significant cost advantages. By compacting the reasoning steps, it reduces the number of input tokens required for few- shot prompting and shortens the output token length, directly lowering computational cost. This token efficiency makes CoD prompting especially appealing in cost- sensitive scenarios, such as large-scale deployments of LLMs or applications with strict budget constraints. CoD prompting also enables fast and efficient reasoning for LLMs. CoD prompting can be combined with other latency-reducing methods, such as adaptive parallel reasoning or multi-pass validation, to further optimize performance across different application domains.

[0088] In some examples, the evaluation engine 440 selects different Al models for performing different types of tasks based on evaluation results, as shown above. For example, the evaluation engine 440 selects the GPT-4o model for performing sports understanding tasks using CoD prompting, because the responses by GPT-4o model show less token count and lower latency while maintaining 100% accuracy using CoD prompting, compared to those by the Claude 3.5 Sonnet model with CoD prompting. The evaluation data, such as the accuracy, token count, and latency, associated with different Al models using different prompts for performing different types of tasks, is store in the data store 410.

[0089] In some examples, the evaluation engine 440 selects certain queries and corresponding generated responses from data store 410, based on corresponding evaluation results, to fine-tune certain generative Al models. Queries and corresponding responses generated by the same generative Al model or another generative Al model whose accuracy is greater than a predetermined accuracy threshold, whose token count is less than a predetermined tokenAtty Docket 1544246--ZM-0880-WOthreshold, or whose latency is less than a predetermined latency threshold can be used to fine-tune the generative Al model. For example, sports understanding tasks and corresponding responses generated by the Claude 3.5 Sonnet model with CoD prompting are selected to fine-tune a GPT-4o model, because the responses by GPT-4o model showed less token count and lower latency while maintaining 100% accuracy using CoD prompting compared to those by the Claude 3.5 Sonnet model with CoD prompting. The communication platform 310 or another component on the communication platform 310 such as a fine-tuning engine (not shown) uses the selected queries and responses to periodically finetune a generative Al model used in the Al-powered response generator 430. The selected queries and responses are training data used for fine-tuning, which can be called training queries and training responses. The training responses include training intermediate steps within the maximum word count.

[0090] The client device 330 is installed with a communication application 450 provided by the communication platform 310. In some examples, the communication application 450 installed on the client device 330 includes a local data store 460, a local model store 470. a local Al-powered response generator 480, and a local evaluation engine 490. The local data store 460 stores communication data generated at or accessible by the client device 330, user data associated with the user of the client device 330, user queries, generated responses, similar to the data store 410 on the communication platform 310. The local model store 470 stores AI / ML models or provides APIs to access AI / ML models on the communication platform 310 or other remote servers 380, similar to the model store 420 on the communication platform 310. The local Al-powered response generator 480 generates a response to a user query using a pre-trained generative Al model configured with an instructional prompt, similar to the Al-powered response generator 430 on the communication platform 310. The local evaluation engine 490 evaluates performances of different pre-trained generative Al models with different instructional prompts, similar to the evaluation engine 440 on the communication platform 310. In some examples, the communication application 450 also includes a graphical user interface (GUI) for the user associated with the client device 330 to enter a user query. The GUI can also display the user query and response generated by the Al-powered response generator 430.

[0091] Referring now to Figure 5, Figure 5 shows an example GUI 500 displaying a consent authorization request for accessing personal data. In someAtty Docket 1544246--ZM-0880-WOexamples according to the present disclosure, a user may select an option to use one or more optional Al features available from a communication platform, such as the chat and video conference provider 110 or the chat and video conference provider 210. The use of these optional Al features may involve providing the user’s personal information to the Al models underlying the Al features. The personal information may include the user’s contacts, calendar, communication histories, video or audio streams, recordings of the video or audio streams, transcripts of audio or video conferences, or any other personal information available the virtual conference provider. Further, the audio or video feeds may include the user’s speech, which includes the user’s speaking patterns, cadence, diction, timbre, and pitch; the user's appearance and likeness, which may include facial movements, eye movements, arm or hand movements, and body movements, all of which may be employed to provide the optional Al features or to train the underlying Al models.

[0092] Before capturing and using any such information, whether to provide optional Al features or to provide training data for the underlying Al models, the user may be provided with an option to consent, or deny consent, to access and use some or all of the user’s personal information. In general, Applicant’s goal is to invest in Al-driven innovation that enhances user experience and productivity while prioritizing trust, safety, and privacy. Without the user’s explicit, informed consent, the user’s personal information will not be used with any Al functionality or as training data for any Al model. Additionally, these optional Al features are turned off by default — account owners and administrators control whether to enable these Al features for their accounts, and if enabled, individual users may determine whether to provide consent to use their personal information.

[0093] As can be seen in Figure 5, a user has engaged in a video conference and has selected an option to use an available optional Al feature. In response, the GUI has displayed a consent authorization window 510 for the user to interact with. The consent authorization window informs the user that their request may involve the optional Al feature accessing multiple different types of information, which may be personal to the user. The user can then decide whether to grant permission or not to the optional Al feature generally, or only in a limited capacity. For example, the user may select an option to only allow the Al functionality to use the personal information to provide the Al functionality.Atty Docket 1544246--ZM-0880-WObut not for training of the underlying Al models. In addition, the user is presented with the option to select which types of information may be shared and for what purpose, such as to provide the Al functionality or to allow use for training underlying Al models.

[0094] Figure 6 shows example responses to the same question generated by the same LLM with three different types of instructional prompts. An example question is “Jim had 20 apples. He gave Dan some apples. Now Jim has 12 apples. How many apples did Jim give to Dan?’' The response generated by an LLM with the base prompt directly outputs the answer, often without any reasoning, as shown in block 602. While correct, this lacks transparency in how the answer was derived, and it requires the language model to run multi-step reasoning without any help from intermediate result, The same LLM with CoT prompting, on the other hand, provides a detailed reasoning process, as shown in block 604. While this response is accurate and interpretable, it includes unnecessary details about Jim, Dan, and the apples, which arc irrelevant to solving the mathematical problem. This verbosity inflates the token count and increases response latency. By contrast, CoT prompting condenses the reasoning process into minimal, abstract representations, as shown in block 606. Here, the reasoning is distilled into a concise equation, focusing solely on the essential mathematical operation required to arrive at the solution. By abstracting away irrelevant contextual details, CoD significantly reduces the token count while maintaining transparency and correctness.

[0095] Now referring to Figure 7, Figure 7 shows an example method for generating a response to a user query using a generative Al model with CoD prompting. At block 702, a communication platform 310 obtains a first instructional prompt for a pre-trained generative artificial intelligence (Al) model, the first instructional prompt comprising an instruction to employ step-by-step reasoning with a maximum word count for each reasoning step. The communication platform 310 includes an Al-powered response generator 430, which uses a pre-trained generative Al model to generate a response to a user query, generally as described in Figure 4. The instructional prompt is predetermined by an administrator user as a system prompt for the generative Al model and can be updated by the administrator user. The instructional prompt includes one or more instructions for step-by-step reasoning with a maximum word count (e.g., 5) for each reasoning step. The instructional promptAtty Docket 1544246--ZM-0880-WOcan also include an instruction of separating a final response from the step-by-step reasoning with a separation mark.

[0096] In some examples, the instructional prompt also includes one or more example queries and corresponding example responses. Each example response includes one or more example intermediate steps and an example final response. Each example intermediate step includes an abstract representation of a natural language. An example query is “Jim had 20 apples. He gave Dan some apples. Now Jim has 12 apples. How many apples did Jim give to Dan?” The corresponding example response is “20 - x = 12; x = 20 — 12. #### 8.” The example intermediate steps are “20 - x = 12” and “x = 20 - 12.”

[0097] At block 704. the communication platform 310 receives a user query for a client device 330. A user associated with the client device 330 enters a user query via a GUI of the Al-powered response generator 430. The user query includes a question or problem for the Al-powered response generator 430 to answer or solve. The user query is a task prompt for the Al-powered response generator 430. In some examples, the user query also includes one or more example queries and example responses. The example queries are similar to the user query, and the example responses show the Al-powered response generator 430 how to respond to the user query.

[0098] At block 706, the communication platform 310 generates a response to the user query using the pre-trained generative Al model based on the first instructional prompt and the user query. The Al-powered response generator 430 on the communication platform 310 uses the pre-trained generative Al model to generate a response to the user query, employing the instructional prompt obtained at block 702. The response includes one or more intermediate steps, each within a maximum word count. The response includes a final response following but separated from the one or more intermediate steps with a separation mark.

[0099] At block 708, the communication platform 310 provides the response to the client device 330. The response generated at block 706 can be displayed via a GUI of the Al-powered response generator 430 following the user query entered by the user at block 704.

[0100] In some examples, an evaluation engine 440 on the communication platform 310 compares responses generated by the pre-trained generative Al model and another pre-trained generative Al model using the first instructionalAtty Docket 1544246-ZM-0880-WOprompt and a second instructional prompt to evaluate the performances of different Al models with different instructional prompts. The evaluation metrics can include accuracy of generated responses, token count in generated responses, and latency of response generation. The evaluation results can guide the communication platform 310 to select proper Al models for the Al-powered response generator 430. Some responses with corresponding user queries, for example those whose accuracy is above a threshold, and token count and latency are below certain thresholds, can be selected based on the evaluation results to fine-tune a pre-trained generative Al model.

[0101] The example method 700 illustrates a method for generating a response to a user query using a generative Al model. However, not every step in the example method 700 may be needed, or some steps may be in a different order. The example method 700 is performed by a client device. Alternatively, the example method 700 can be performed by a communication platform or any suitable servers that can generate a response to a user query using a generative Al model.

[0102] Referring now to Figure 8, Figure 8 shows an example computing device 800 suitable for use in example systems or methods of generating a response to a user query using a generative Al model according to this disclosure. The example computing device 800 includes a processor 810 which is in communication with the memory 820 and other components of the computing device 800 using one or more communications buses 802. The processor 810 is configured to execute processor-executable instructions stored in the memory 820 to perform one or more methods of generating a response to a user query using a generative Al model according to different examples, such as part or all of the example method 700 described above with respect to Figure 7. In some embodiments, the computing device may include software 860 for executing one or more methods described herein, such as for example, one or more steps of method 700. The computing device 800, in this example, also includes one or more user input devices 850, such as a keyboard, mouse, touchscreen, microphone, etc., to accept user input. The computing device 800 also includes a display 840 to provide visual output to a user.

[0103] In addition, the computing device 800 includes software 860 to enable a user to join and participate in one or more virtual spaces or in one or more conferences, such as a conventional conference or webinar, by receivingAtty Docket 1544246-ZM-0880-WOmultimedia streams from a virtual conference provider, sending multimedia streams to the virtual conference provider, joining and leaving breakout rooms, creating video conference expos, etc., such as described throughout this disclosure, etc.

[0104] The computing device 800 also includes a communications interface 830. In some examples, the communications interface 830 may enable communications using one or more networks, including a local area network (“LAN’ ); wide area network (“WAN”), such as the Internet; metropolitan area network (“MAN”); point-to-point or peer-to-peer connection; etc. Communication with other devices may be accomplished using any suitable networking protocol. For example, one suitable networking protocol may include the Internet Protocol (“IP”), Transmission Control Protocol (“TCP”), User Datagram Protocol (“UDP”), or combinations thereof, such as TCP / IP or UDP / IP.

[0105] While some examples of methods and systems herein are described in terms of software executing on various machines, the methods and systems may also be implemented as specifically configured hardware, such as field-programmable gate array (FPGA) specifically to execute the various methods according to this disclosure. For example, examples can be implemented in digital electronic circuitry, or in computer hardware, firmware, software, or in a combination thereof. In one example, a device may include a processor or processors. The processor comprises a computer-readable medium, such as a random-access memory (RAM) coupled to the processor. The processor executes computer-executable program instructions stored in memory, such as executing one or more computer programs. Such processors may comprise a microprocessor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), field programmable gate arrays (FPGAs), and state machines. Such processors may further comprise programmable electronic devices such as PLCs, programmable interrupt controllers (PICs), programmable logic devices (PLDs), programmable read-only memories (PROMs), electronically programmable readonly memories (EPROMs or EEPROMs), or other similar devices.

[0106] Such processors may comprise, or may be in communication with, media, for example one or more non- transitory computer-readable media, that may store processor-executable instructions that, when executed by the processor, can cause the processor to perform methods according to this disclosure as carried out, or assisted, by a processor. Examples of non-transitoryAtty Docket 1544246-ZM-0880-WOcomputer-readable medium may include, but are not limited to, an electronic, optical, magnetic, or other storage device capable of providing a processor, such as the processor in a web server, with processor-executable instructions. Other examples of non-transitory computer-readable media include, but are not limited to, a floppy disk, CD-ROM, magnetic disk, memory chip, ROM. RAM, ASIC, configured processor, all optical media, all magnetic tape or other magnetic media, or any other medium from which a computer processor can read. The processor, and the processing, described may be in one or more structures, and may be dispersed through one or more structures. The processor may comprise code to carry out methods (or parts of methods) according to this disclosure.

[0107] The foregoing description of some examples has been presented only for the purpose of illustration and description and is not intended to be exhaustive or to limit the disclosure to the precise forms disclosed. Numerous modifications and adaptations thereof will be apparent to those skilled in the art without departing from the spirit and scope of the disclosure.

[0108] Reference herein to an example or implementation means that a particular feature, structure, operation, or other characteristic described in connection with the example may be included in at least one implementation of the disclosure. The disclosure is not restricted to the particular examples or implementations described as such. The appearance of the phrases “in one example,” “in an example.” “in one implementation.” or “in an implementation,” or variations of the same in various places in the specification does not necessarily refer to the same example or implementation. Any particular feature, structure, operation, or other characteristic described in this specification in relation to one example or implementation may be combined with other features, structures, operations, or other characteristics described in respect of any other example or implementation.

[0109] Use herein of the word “or” is intended to cover inclusive and exclusive OR conditions. In other words, A or B or C includes any or all of the following alternative combinations as appropriate for a particular usage: A alone; B alone; C alone; A and B only; A and C only; B and C only; and A and B and C.

Claims

1. Atty Docket 1544246-ZM-0880-WOCLAIMSThat which is claimed is:

1. A method comprising:obtaining a first instructional prompt for a pre-trained generative artificial intelligence (Al) model, the first instructional prompt comprising an instruction to employ step-by-step reasoning with a maximum word count for each reasoning step;receiving a user query from a client device;generating a response to the user query using the pre-trained generative Al model based on the first instructional prompt and the user query; and providing the response to the client device.

2. The method of claim 1, wherein the first instructional prompt further comprises an example query and an example response, and wherein the example response comprises one or more example intermediate steps and an example final response.

3. The method of claim 2, wherein each of the one or more example intermediate steps comprises an abstract representation of a natural language.

4. The method of claim 1. further comprising fine-tuning the pre-trained generative Al model with training data comprising training intermediate steps within the maximum word count.

5. The method of claim 1, wherein the first instructional prompt further comprises an instruction of separating a final response from the step-by- step reasoning with a separation mark.

6. The method of claim 1, further comprising:obtaining a second instruction prompt: andgenerating a second response to the user query using the pre-trained generative Al model based on the second instructional prompt and the user query.

7. The method of claim 1, further comprising:Atty Docket 1544246-ZM-0880-WOevaluating a first performance of the pre-trained generative Al model generating the first response based on the first instructional prompt and a second performance of the pre-trained generative Al model generating the second response based on the second instructional prompt to generate evaluation data, wherein the evaluation data comprises accuracy, token count, or latency.

8. A system comprising:a communications interface;a non- transitory computer-readable medium; andone or more processors communicatively coupled to the communications interface and the non-transitory computer-readable medium, the one or more processors configured to execute processor-executable instructions stored in the non-transitory computer-readable medium to:obtain a first instructional prompt for a pre-trained generative artificial intelligence (Al) model, the first instructional prompt comprising an instruction to employ step-by-step reasoning with a maximum word count for each reasoning step;receive a user query from a client device;generate a response to the user query using the pre-trained generative Al model based on the first instructional prompt and the user query; and provide the response to the client device.

9. The system of claim 8, wherein the first instructional prompt further comprises an example query and an example response, and wherein the example response comprises one or more example intermediate steps and an example final response.

10. The system of claim 9, wherein each of the one or more example intermediate steps comprises an abstract representation of a natural language.

11. The system of claim 8. wherein the one or more processors are configured to execute further processor-executable instructions stored in the non- transitory computer-readable medium to:fine-tune the pre-trained generative Al model with training data comprising training intermediate steps within the maximum word count.Atty Docket 1544246-ZM-0880-WO12. The system of claim 8, wherein the first instructional prompt further comprises an instruction of separating a final response from the step-by- step reasoning with a separation mark.

13. The system of claim 8. wherein the one or more processors are configured to execute further processor-executable instructions stored in the non- transitory computer-readable medium to:obtain a second instruction prompt; andgenerate a second response to the user query using the pre-trained generative Al model based on the second instructional prompt and the user query.

14. The system of claim 8. wherein the one or more processors are configured to execute further processor-executable instructions stored in the non- transitory computer-readable medium to:evaluate a first performance of the pre-trained generative Al model generating the first response based on the first instructional prompt and a second performance of the pre-trained generative Al model generating the second response based on the second instructional prompt to generate evaluation data, wherein the evaluation data comprises accuracy, token count, or latency.

15. A non-transitory computer-readable medium comprising processorexecutable instructions configured to cause one or more processors to: obtain a first instructional prompt for a pre-trained generative artificial intelligence (Al) model, the first instructional prompt comprising an instruction to employ step-by-step reasoning and a maximum word count for each reasoning step;receive a user query from a client device;generate a response to the user query using the pre-trained generative Al model based on the first instructional prompt and the user query; and provide the response to the client device.

16. The non-transitory computer-readable medium of claim 15, wherein the first instructional prompt further comprises an example query and an example response, and wherein the example response comprises one or more example intermediate steps and an example final response.Atty Docket 1544246-ZM-0880-WO17. The non-transitory computer-readable medium of claim 16, wherein each of the one or more example intermediate steps comprises an abstract representation of a natural language.

18. The non-transitory computer-readable medium of claim 15. further comprising processor-executable instructions configured to cause one or more processors to:fine-tune the pre-trained generative Al model with training data comprising training intermediate steps within the maximum word count.

19. The non-transitory computer-readable medium of claim 15. further comprising processor-executable instructions configured to cause one or more processors to:obtain a second instruction prompt; andgenerate a second response to the user query using the pre-trained generative Al model based on the second instructional prompt and the user query.

20. The non-transitory computer-readable medium of claim 19, further comprising processor-executable instructions configured to cause one or more processors to:evaluate a first performance of the pre-trained generative Al model generating the first response based on the first instructional prompt and a second performance of the pre-trained generative Al model generating the second response based on the second instructional prompt to generate evaluation data, wherein the evaluation data comprises accuracy, token count, or latency.

Images