Abstract: A testing environment provisioning service obtains configuration information for an AI application based on monitoring its deployment environment to determine an initial configuration thereof. Configuration information included in the initial configuration includes configuration information for the language model(s) used by the AI application, a prompt template(s) used by the AI application, and a configuration of a content filter that interfaces with the language model(s). The service monitors the deployment environment for changes to this initial configuration by periodically obtaining additional configuration information for the deployment environment and evaluating the additional configuration information against initial configuration. When a substantial change to the initial configuration is detected, the testing service updates the initial configuration and provisions a testing environment with a language model(s) and any content filter configured according to the updated configuration.

Abstract: Provided are anti-FGFR2b antibodies or antigen-binding fragments thereof, isolated polynucleotides encoding the same, pharmaceutical compositions comprising the same, and the uses thereof.

Abstract: A semiconductor device, a memory, a system and a method of data processing are disclosed in the present disclosure. An example semiconductor device includes a logic die and a plurality of memory dies which are stacked, the logic die comprises a plurality of data processing units each comprising a serial-parallel conversion circuit and an interface circuit. The interface circuit is configured to output the second readout data to a processor and receive first write data with a bit width of N from the processor and transmit the first write data to the serial-parallel conversion circuit. The serial-parallel conversion circuit is configured to parallelize the first write data to obtain second write data with a bit width of M and transmit the second write data to the at least one of the memory dies, wherein M and N are positive integers, and M/N is greater than or equal to 2.

Abstract: A testing framework conducts a dialogue with a multi-modal foundation model to determine how to interact with an application via a user interaction emulator tool based on images captured from the UI of the application. Initially, the testing framework will capture an image of the application UI. The testing framework then prompts the foundation model to explain how to interact with the application to achieve a function of the application given the captured image. The testing framework derives commands from the model response for running the user interaction emulator tool and captures an image of the application UI after the interaction and again prompts the foundation model to determine whether the function was successfully performed or to explain a failure. The testing framework continues the dialogue with successive prompts for interaction with the application informed by the explanation of preceding failure.

Abstract: Various techniques for LLM powered detection reasoning solutions are disclosed. In some embodiments, a system, a process, and/or a computer program product for an LLM powered detection reasoning solution includes monitoring network traffic at a security platform, wherein the security platform generates a sample based on the monitored network traffic; sending the sample to a security service to generate a Large Language Model (LLM) powered detection and reason, wherein the LLM is prompted to automatically generate a malware or benign verdict and a reason for explaining the verdict; and reporting the LLM powered detection and reason.

Abstract: A gas-liquid separation device includes a first cylinder, a second cylinder, a heat exchange assembly and a gas-liquid separation assembly. The gas-liquid separation device defines a first cavity and a second cavity. The first cavity includes a space between the first cylinder and the second cylinder. The second cavity includes an inner cavity of the second cylinder. The gas-liquid separation assembly is at least partially located in the second cavity. An inner cavity of the gas-liquid separation assembly is in communication with the first cavity and the second cavity. At least part of the heat exchange assembly is located in the first cavity. The gas-liquid separation device includes a first pipe portion. A pipe cavity of the first pipe portion communicates with the second cavity and an outer space of the first cylinder. A thermal management system having the gas-liquid separation device is also disclosed.

Abstract: A prompt reinforcement tool (“tool”) compares benign sample prompts to determine common and differing content across the prompts. The tool prompts a LLM to identify system prompt components from the common prompt content and identify user inputs and additional inputs from the differing prompt content. The tool also prompts the LLM to extract prompt components based on the response samples. The prompt reinforcement tool updates an application agnostic prompt template and generates an application tailored prompt template. To update, the tool removes irrelevant sections and populates remaining/preserved sections of the application agnostic prompt template that correspond to the identified system prompt components and the extracted prompt components. The tool then inserts, adjacent to template sections for user input and additional input, placeholders for a secure wrapper and inserts a guardrail instruction to isolate any commands or instructions that may be in the user input or additional input.

Abstract: A security feed normalizer aggregates and normalizes threat intelligence data across security feeds and extracts threat descriptors of cybersecurity threats from the aggregated/normalized data. A first large language model (LLM) determines whether each threat descriptor is informative, i.e., comprises sufficient information for reproducing/generating network traffic of the corresponding cybersecurity threat. For informative threat descriptors, a second LLM generates network traffic for the corresponding cybersecurity threats. The generated network traffic is used for subsequent remediation of corresponding threats.

Abstract: A system facilitates multi-session, multi-aspect analysis of conversations between users and applications that leverage language models (“generative AI applications”). The system retrieves conversation logs for conversations carried out between users and a generative AI application(s) and sorts the conversation logs by conversation. The system analyzes the conversations in multiple aspects for various use cases: security, analytics, and organizational compliance. In the security analysis, the system analyzes the queries and responses of each conversation to determine if the conversation includes a multi-turn attack. For the analytics and organization compliance use cases, the system evaluates how end users are using the generative AI application(s) across conversations to provide customers with additional insights, such as the spread of subject matter areas for which employees are using the generative AI application(s).

Abstract: A genetic algorithm is implemented to generate prompts that evade content filters of generative artificial intelligence (AI) systems. The genetic algorithm applies grammar operations to mutate candidate prompts, communicates the candidate prompts to generative AI systems, and selects candidate prompts that successfully evade content filters according to corresponding responses. A disambiguation model that corrects grammar in prompts is tested on the selected prompts to determine if grammar is properly corrected. Once tested, the disambiguation model is deployed in an ensemble with a classifier that outputs verdicts for prompts with grammar corrected by the disambiguation model.

Abstract: Techniques for detecting anomalous network behavior in operational technology (OT) protocols are disclosed. A system, process, and/or computer program product for detecting anomalous network behavior in OT protocols include monitoring network traffic to perform automated OT malware detection analysis of OT related network traffic, extracting one or more features from the OT related network traffic, inputting the one or more extracted features into a model for malware detection analysis, and performing an action based on a result of the model.

Abstract: An air conditioner control method, an air conditioner, and a storage medium. The air conditioner is provided with the detection apparatus group. The method includes controlling the detection apparatus group to be in different detection modes to detect a state of a target object; and controlling operation of the air conditioner based on the state of the target object.

Abstract: Provided herein are compounds of formula A-B, where A is selected from a monosaccharide or a disaccharide and B is a therapeutic moiety (e.g. selected from an anti-inflammatory moiety or chemotherapeutic moiety); as well as a pharmaceutically acceptable salt, stereoisomer, acid, or solvate thereof. Such a compound may be a substrate for glucosylceramidase beta 2 (GBA2), such that the compound is a prodrug that provides an active compound after cleavage of the monosaccharide or disaccharide by GBA2. Also provided are pharmaceutical compositions comprising the compounds, as well as various medical and non-medical uses of the compounds and compositions.

Abstract: The present invention provides a VOPRF protocol for post-quantum security, and a method and system for anonymous token authentication. The VOPRF protocol includes: a public parameter setting phase, an encoding and request phase, a re-encoding and response phase, and a decoding phase. The method for anonymous token authentication includes: an initialization phase: generating, by a trusted third-party, a common reference string; a key generation phase; a signing phase, executing jointly, by a client and a server, the VOPRF protocol to generate an anonymous token; and a token redemption phase: extracting, by the client, a token from a local list, and sending the token to the server for verification. The present invention implements the VOPRF protocol for post-quantum security and the method and system for anonymous token authentication, to still ensure security of an algorithm and a protocol under the threat of strong computing power of a quantum computer.

Abstract: Disclosed are a three-dimensional modeling system and modeling method based on multimodal fusion. The method includes: separately collecting feedback data of an electroencephalogram sensor, an electromyography sensor, an eye movement sensor, a gesture sensor, and a voice sensor, conducting multimodal fusion on the feedback data, obtaining multimodal-fused model data, matching the model data with a database instruction, obtaining and analyzing an instruction set, obtaining and identifying a relevant modeling parameter, obtaining a modeling method, automatically conducting modeling according to the modeling method, and obtaining a visual entity model.

Abstract: Various techniques for LLM powered detection reasoning solutions are disclosed. In some embodiments, a system, a process, and/or a computer program product for an LLM powered detection reasoning solution includes monitoring network traffic at a security platform, wherein the security platform generates a sample based on the monitored network traffic; sending the sample to a security service to generate a Large Language Model (LLM) powered detection and reason, wherein the LLM is prompted to automatically generate a malware or benign verdict and a reason for explaining the verdict; and reporting the LLM powered detection and reason.

Abstract: A security feed normalizer aggregates and normalizes threat intelligence data across security feeds and extracts threat descriptors of cybersecurity threats from the aggregated/normalized data. A first large language model (LLM) determines whether each threat descriptor is informative, i.e., comprises sufficient information for reproducing/generating network traffic of the corresponding cybersecurity threat. For informative threat descriptors, a second LLM generates network traffic for the corresponding cybersecurity threats. The generated network traffic is used for subsequent remediation of corresponding threats.

Abstract: A prompt generator receives descriptive data for a plug-in of a custom or open-source security testing tool. The prompt generator chooses a prompt template corresponding to the type of descriptive data and the security testing tool. For an open-source security testing tool, the prompt generator populates the prompt template with the descriptive data and prompts an open-source large language model (LLM) with the populated prompt to generate the plug-in for the open-source security testing tool. For a custom security testing tool, the prompt generator populates the prompt template with the descriptive data or an existing open source plugin, and prompts a custom LLM with the populated prompt to generate the plug-in for the custom security testing tool.

Abstract: According to one aspect of the present disclosure, a memory apparatus is provided. The memory apparatus may include a first memory die and a die group stacked in a first direction. The die group may include M second memory dies stacked in the first direction. Each of the second memory dies may be connected to a third memory die and configured to perform data exchange at a bandwidth of (N*X) bits with the third memory die through N data channels. The first memory die may be connected to the third memory die. The first memory die may be configured to perform data exchange at a bandwidth of (N*X) bits with the third memory die through N data channels. The first memory die may be configured to replace a first data channel of the die group when a storage portion corresponding to the first data channel fails.

Abstract: Disclosed is an intelligent interactive decision-making method for a discrete manufacturing system. The method includes the following steps: step 1, establishing a production scheduling optimization model and strategy for discrete manufacturing for an actual application scene; step 2, training the scheduling strategy with existing production data on the basis of a deep reinforcement learning algorithm, and storing a state having a high reward in a training process in a memory; step 3, updating the state according to prior knowledge in the memory; step 4, inputting the updated state into a deep reinforcement learning network, obtaining a corresponding reward, and updating the memory according to the reward; and step 5, repeating step 4 until model parameters converge, and saving and putting the model into an actual production scene.