21 results about "Perplexity" patented technology

Embodiments herein describe a sparse-dense-sparse (SDS) process that achieves a better pruning scheme that benefits from pruning-friendliness relative to one-shot pruning schemes. The SDS process performs a first pruning to generate a sparse ML model followed by reconstruction to generate a re-dense ML model, followed by a second pruning to generate another sparse ML model. By pruning a ML model and then re-constructing the ML model, the ML model can be made more pruning-friendly by performing data and / or weight regularization. As a result, performing the second pruning in the SDS process can result in a smoother weight distribution and lower perplexity relative to one-shot pruning.

The application discloses a kind of based on big language model's evolvable software vulnerability detection method, comprising: by regular pattern matching identification Source sentence, based on call graph traversal and data dependence analysis execution function level inter-process slice, build cross-function code context, input the big language model of parameter efficient fine-tuning, output the vulnerability propagation path from Source to Sink;When new vulnerability type needs to be extended, the parameter variation characteristics of old data are extracted by multi-step fine-tuning, and the representative core set is selected by random projection dimension reduction and hybrid distance hierarchical clustering, and the training is played back by mixing new data to alleviate catastrophic forgetting;In the actual use process of tool, the false alarm and the false alarm confirmed by user are collected as feedback signal, the core set is clustered and layered filtered and refined based on perplexity and error prediction analysis, the harmful old knowledge that leads to false alarm and false alarm is removed, and verified new mode is supplemented at the same time, to realize the closed-loop evolution of self-improvement.

A system for influencing outputs of a large language model includes at least one memory, at least one processor, a perplexity optimizer, and a corpus handler. The processor executes instructions stored in the memory to operate the optimizer and handler. The perplexity optimizer generates multiple candidate supporting texts based on a target concept. It computes a perplexity metric for each candidate within a context derived from a local corpus, using token likelihoods from a reference language model, and selects a supporting text based on the metric. The corpus handler identifies online editable corpora based on their likelihood of inclusion in language model training data. It then collects contextual information from these corpora to create the local corpus and inserts the selected supporting text into at least one of the identified online corpora.

The application provides a large language model individualization alignment method and device based on difficulty perception learning, and belongs to the technical field of model training. The method comprises the following steps: obtaining multiple groups of training samples; each group of training samples comprises a user question, a user portrait, and a target reply corresponding to the user question under the user portrait; calculating the perplexity of the target reply in each group of training samples; dividing the multiple groups of training samples into high-perplexity training samples and low-perplexity training samples according to a preset perplexity threshold and the perplexity of the target reply in each group of training samples; performing first training on a preset large language model based on the low-perplexity training samples and a likelihood loss function, performing second training on the preset large language model after the first training based on the high-perplexity training samples and reinforcement learning, and obtaining a target large language model, so that the target large language model completes individualization alignment. The application can improve the individualization alignment effect of the model.

This invention presents an AI-generated text detection method based on graph structure features, belonging to the fields of artificial intelligence and natural language processing. The method includes: dataset construction, entity relation extraction and graph structure construction, graph structure feature extraction, graph structure feature model training, and text detection. It further incorporates traditional text feature extraction and model training, adaptively fusing the traditional text feature model and the graph feature model based on confidence-weighted entropy, and then performing text detection based on the fused model. This invention is the first to perform AI text detection from the perspective of graph structure features, breaking through the limitations of existing research that focuses on surface features such as vocabulary, syntax, and perplexity. The fusion strategy dynamically adjusts the fusion weights by quantifying the uncertainty of model predictions, maintaining a high level of performance on both original data and adversarial examples, achieving a balance between detection accuracy and adversarial robustness. It can be widely applied to the detection of AI-generated content such as news content and academic papers.

Method and device for detecting intrusion in a computer system. This method uses a large language model (LLM), previously trained by machine learning on training data comprising at least training data representative of a normal state of network communications or of the operation of the computer system, and comprises the steps of: slicing (42) at least a subset of collected formatted data into a sequence of elementary fragments, applying (44) the LLM to the sequence of elementary fragments, providing as output, for each elementary fragment, a likelihood value of the elementary fragment as a function of a context comprising at least some of the other elementary fragments of said sequence of elementary fragments, calculating (46) a perplexity value of said sequence of elementary fragments as a function of said likelihood values,When the calculated perplexity value exceeds a normality threshold, detection (50) of a potential intrusion and issuance of an intrusion alert in said computer system. Figure for the abbreviation: Figure 2,

This invention discloses an automatic mining and classification method for private data, comprising the following steps: First, receiving private unstructured documents, performing preprocessing and semantic alignment, and using sliding window technology to segment the documents into continuous text blocks; then loading a locally deployed general basic model and an industry-specific model, calculating the perplexity of each text block, and determining the scarcity of the text block by comparing the differences in the model output; next, constructing a statistical model based on the perplexity distribution of the enterprise's historical documents and dynamically setting a threshold, calculating the value score for text blocks exceeding the threshold; finally, projecting the scarcity and value score onto a preset classification decision matrix to automatically determine the secret level of the text block and trigger corresponding security handling strategies. This invention achieves automated and intelligent classification of private data, improving the efficiency and accuracy of data security management.

This invention relates to a method for constructing a large-scale Burmese-Chinese parallel corpus based on multi-step thinking. The invention includes: translating existing Chinese-English parallel corpora using currently available translation models to obtain original English-Burmese and Chinese-Burmese parallel sentence pairs; calculating double-confidence intervals for semantic similarity and perplexity between aligned Burmese-Chinese sentence pairs based on publicly available high-quality Burmese-Chinese parallel corpora; using the selected double-confidence intervals to perform preliminary screening of Burmese-Chinese parallel sentence pairs, forming pre-processed pseudo-parallel sentence pairs; designing a multi-step thinking chain to guide the large-scale model to progressively optimize the pre-processed pseudo-parallel sentence pairs, thereby generating high-quality Burmese-Chinese parallel corpora for training the translation model, thus effectively improving the performance of Burmese-Chinese machine translation. This invention significantly enhances the ability of large language models to construct corpora in Burmese, a low-resource language, and provides an interpretable and transferable technical paradigm for corpus construction in other low-resource languages.

Certain aspects of the disclosure provide a method for training a language model (LM) including: generating, using an LM, one or more outputs; computing a confidence score of an output of the one or more outputs based on a perplexity value of the output; determining, by a group relative policy optimization (GRPO)-based model, that the output is: associated with a correct status based on a reference policy; and associated with an uncertain status based on the confidence score and a threshold; determining, by the GRPO-based model, an increased reward value for the output that is associated with the correct status and the uncertain status based at least in part on a base reward value and the confidence score; causing, by the GRPO-based model, a reinforcement of the output using the increased reward value; and training the LM in accordance with the reinforcement of the output.

This invention provides a method and system for assessing bias risk in context learning based on the construction and injection of Chinese text style. The method includes: constructing a definition of Chinese text style; building a style feature injector based on the most sensitive Chinese text style; injecting style features into the input portion of examples in the context learning prompt template to form stylized example input; synthesizing biased output using a teacher model, and replacing the examples in the context learning prompt template with the stylized example input and biased output; calling the evaluated large language model to perform original user queries and stylized user queries respectively to generate model output; using the Regard index and bias susceptibility rate to quantitatively analyze the model output, and using perplexity and semantic similarity as quality control indicators to obtain the bias risk assessment result. This invention reveals the bias risk of large language models in context learning scenarios and improves the effectiveness and stability of bias risk assessment.

The present application provides a large model endogenous security defense method based on interlayer contrast decoding, belonging to the technical field of artificial intelligence and network security, which can at least partially solve the problems of high fine-tuning cost, lagging defense and easy illusion of the existing large language model security defense means. The present application comprises: offline construction of field risk data set and input of model, utilization of JS divergence to analyze the difference between each layer output and final layer output, positioning of high-risk sensitive layer set and extraction of harmful feature fingerprint; real-time monitoring of the activation mode of generated word units in the sensitive layer in the inference stage, triggering of security intervention when the similarity with the harmful feature fingerprint exceeds the threshold; execution of negative contrast decoding, stripping of the sensitive layer output from the final layer output through weighted subtraction operation; introduction of an adaptive rollback mechanism, dynamic adjustment of the inhibition coefficient or rollback decoding mode according to the perplexity. The present application can dynamically block hidden semantic attack and harmful content generation at the decoding end without fine-tuning model parameters.

The application discloses a kind of contextual adaptive interaction evaluation method and system, it is related to human-computer interaction evaluation, data processing and artificial intelligence technical field, in acquisition end gathers region information, main use language, education years, historical evaluation record and compliance record to generate cultural context embedding;In side presentation contextual interaction topic and interaction task and gather answer duration, prompt number, error type, touch or mouse track, voice response and image or image data;According to perplexity and answer behavior trigger alternative question state machine and calculate rectification cost;Based on external objective calibration index, scale or task score to consistency score monotone mapping is established, and the consistency score is used as fusion anchor quantity;For each mode, programmable expert output layered judgment result and uncertainty are set, to improve the consistency and robustness of evaluation in cross-region, cross-language and low-education population scenarios.

This application relates to the fields of artificial intelligence and natural language processing technology, and discloses a method and system for significantly reducing the AI ​​features of text generated by large models. It aims to solve the problems of existing technologies having rigid rewriting logic, easily disrupting semantic coherence, and lacking closed-loop protection. The method includes: scanning the perplexity and burstiness indicators of the text to be processed to generate a feature distribution heatmap; assigning a reconstruction engine and dynamically configuring weights by a scheduling center; using a semantic fractal architecture to deconstruct the text into atomic-level units, performing parallel human-like reconstruction and semantic fitting under global constraints; and performing consistency and probability checks through multi-level audits, triggering recursive self-healing optimization if the standards are not met. The system includes a feature analysis module, a scheduling center, a parallel reconstruction matrix, and a closed-loop audit module. This application achieves accurate identification and closed-loop repair of AI features, significantly improving the naturalness, concealment, and logical rigor of the generated text while implementing hard constraint protection for core information.

A system for influencing outputs of a large language model includes at least one memory, at least one processor, a perplexity optimizer, and a corpus handler. The processor executes instructions stored in the memory to operate the optimizer and handler. The perplexity optimizer generates multiple candidate supporting texts based on a target concept. It computes a perplexity metric for each candidate within a context derived from a local corpus, using token likelihoods from a reference language model, and selects a supporting text based on the metric. The corpus handler identifies online editable corpora based on their likelihood of inclusion in language model training data. It then collects contextual information from these corpora to create the local corpus and inserts the selected supporting text into at least one of the identified online corpora.

This application relates to the field of large language model technology, providing a model routing method and apparatus, electronic device, and program product. The method includes: determining the perplexity of a first micro-language model deployed locally for each token included in a query request; obtaining a first difficulty index based on the perplexity; counting the first number of predefined features in the query request using regular expression matching; obtaining a second difficulty index based on the number of tokens included in the query request and the first number; obtaining a semantic abstraction score corresponding to the query request as a third difficulty index; obtaining a difficulty score value corresponding to the query request based on the first, second, and third difficulty indices; determining a set of target candidate models corresponding to the query request based on the difficulty score value and a preset difficulty threshold; and selecting a candidate model from the set of target candidate models as the target model. This method can improve the rationality of model routing.

The application discloses a data filtering method for enhancing semantic stability of large model fine-tuning data, and belongs to the technical field of natural language processing, and comprises the following steps: S1, a teacher model is used to process a seed instruction set to generate initial response content; S2, according to the initial response content, a teacher model perplexity and a student model perplexity are generated; S3, according to the teacher model perplexity and the student model perplexity, a value quadrant is determined; S4, based on the initial response content, the value quadrant is screened; and S5, based on the screened value quadrant, a fine-tuning training set of the student model is generated. Through the innovative double-model perplexity residual error calculation and semantic stability calibration mechanism, the application realizes significant technical progress and good technical effects in multiple dimensions.

This invention provides an intelligent operation and maintenance fault diagnosis method based on problem reconstruction and consensus reasoning. The method includes: constructing a fault description rewriting model based on Direct Preference Optimization (DPO) by leveraging the generation capabilities and self-evaluation feedback of a large language model; reconstructing the original fault description in diverse ways based on the fault description rewriting model, and filtering non-standard fault descriptions into a set of fault query statements through a rejection sampling mechanism based on perplexity level (PPL); performing multi-perspective hybrid consensus diagnostic decision-making based on the real-time collected fault descriptions and fault query statement set, and outputting intelligent operation and maintenance fault diagnosis results corresponding to the real-time collected fault descriptions, including root cause localization and remediation suggestions. This invention uses the PPL rejection sampling mechanism to filter high-quality candidate rewrites that conform to operation and maintenance specifications, and utilizes a hybrid majority voting strategy to eliminate comprehension biases from a single perspective, achieving highly robust automated fault localization and diagnosis.

The application belongs to the field of natural language processing, and particularly relates to a text blocking method based on double-feature fusion; the method comprises the following steps: cleaning, standardizing and sentence splitting of original text to obtain text to be blocked; normalization and weighted fusion based on perplexity increment features and semantic inconsistency features of adjacent sentences to determine comprehensive scores of blocking boundaries and generate an initial text block set; backtracking splitting, adjacent block merging and boundary overlap optimization of the initial text block set to obtain a final text block set; the application can improve the boundary accuracy, semantic integrity and logical coherence of text blocking by fusing perplexity increment features and semantic inconsistency features.