Methods and apparatus for selecting supervised fine-tuning training data for person-job matching models

Abstract

This invention discloses a method and apparatus for selecting supervised fine-tuning training data for a job-person matching model, relating to the field of human resource data processing technology based on artificial intelligence. The method includes: obtaining an initial difficulty score for job-resume pairs through multi-dimensional difficulty assessment; generating matching labels based on the actual behavior of recruitment decision-makers; obtaining a comprehensive difficulty score by calibrating the consistency between the model's initial judgment and the matching labels; and constructing a training set with a specific inference style based on data appropriate to the comprehensive difficulty score. This invention achieves precise screening and optimized configuration of training data, effectively improving the accuracy, generalization ability, and training efficiency of the job-person matching model.

Description

Technical Field

[0001] This invention relates to the field of human resource data processing technology based on artificial intelligence, and in particular to a method and apparatus for selecting supervised fine-tuning training data for a person-job matching model. Background Technology

[0002] In the current AI-driven recruitment technology field, job-person matching models based on natural language processing have become a core tool for improving the efficiency of resume screening and job recommendation. The performance of these models, especially in the supervised fine-tuning stage, is highly dependent on the quality of the training data used. Existing technologies have the following main shortcomings when constructing supervised fine-tuning training sets:

[0003] (1) The inherent difficulty difference in matching judgment between different job descriptions and resume combinations was ignored. Simple matching tasks (such as direct comparison of education and years of work experience) and complex matching tasks (such as deep reasoning of skill transferability and achievement correlation) were used in combination, which led to the model overfitting to simple patterns and underfitting to complex cases, thus restricting the model's generalization ability and robustness.

[0004] (2) Lack of a closed-loop business feedback mechanism. The labeling of training data is usually based on offline static labels, which do not incorporate the real behavioral feedback of recruitment decision-makers (such as corporate HR), resulting in a disconnect between data selection and actual business needs, making it difficult for the model to learn key decision factors.

[0005] (3) It cannot guide the construction of the training set. There is a specific synergistic effect between reasoning process data of different styles and source data of different judgment difficulty, but existing methods have not systematically analyzed this relationship, resulting in blind selection of training data and limiting the further improvement of model training efficiency and effect.

[0006] Therefore, there is an urgent need for a new technology that can accurately assess the difficulty of matching job descriptions with candidate resumes, integrate real business feedback for dynamic calibration, and thus scientifically guide the selection of training data and the construction of training sets during the fine-tuning phase. Summary of the Invention

[0007] In view of the above-mentioned defects or deficiencies in the prior art, the present invention provides a method and apparatus for selecting supervised fine-tuning training data for a person-job matching model, which can effectively solve all the technical problems mentioned in the background art.

[0008] One aspect of the present invention provides a method for selecting supervised fine-tuning training data for a person-job matching model, comprising the following steps:

[0009] The difficulty of matching job description-candidate resume pairs is evaluated from multiple dimensions using an artificial intelligence model. A difficulty score is obtained for each dimension, and the difficulty scores for each dimension are weighted and summed to obtain the initial difficulty score for the job description-candidate resume pair.

[0010] Obtain matching tags generated based on the actual behavior of recruitment decision-makers, including positive and negative feedback actions on candidate resumes; use an artificial intelligence model to perform preliminary matching judgment on the job description-candidate resume pair to obtain preliminary matching results. If the preliminary matching result is consistent with the matching tags, the initial difficulty score is downweighted; if the preliminary matching result is inconsistent with the matching tags, the initial difficulty score is upweighted to obtain the comprehensive difficulty score of the job description-candidate resume pair.

[0011] Based on the comprehensive difficulty score, a supervised fine-tuning training set for a specific reasoning style is constructed by selecting job descriptions and candidate resumes that match the difficulty level; wherein, the reasoning style is supervised fine-tuning training data that differs in the level of detail of the reasoning process, logical structure, or language style.

[0012] In another aspect, the present invention provides a device for selecting supervised fine-tuning training data for a person-job matching model, comprising:

[0013] The model scoring module is used to evaluate the difficulty of matching job description-candidate resume pairs in multiple dimensions using an artificial intelligence model, obtain a difficulty score for each dimension, and then perform a weighted summation of the difficulty scores for each dimension to obtain the initial difficulty score for the job description-candidate resume pair.

[0014] The scoring calibration module is used to obtain matching tags generated based on the actual behavior of recruitment decision-makers, including positive and negative feedback operations on candidate resumes; the job description-candidate resume pair is initially matched and judged by an artificial intelligence model to obtain an initial matching result; if the initial matching result is consistent with the matching tag, the initial difficulty score is downweighted; if the initial matching result is inconsistent with the matching tag, the initial difficulty score is upweighted to obtain a comprehensive difficulty score for the job description-candidate resume pair.

[0015] The training data filtering module is used to select job descriptions and candidate resumes that match the difficulty level based on the comprehensive difficulty score to construct a supervised fine-tuning training set for a specific reasoning style; wherein, the reasoning style is supervised fine-tuning training data that is differentiated in terms of the level of detail of the reasoning process, logical structure, or language style.

[0016] The method and apparatus for selecting supervised fine-tuning training data for the person-job matching model provided by this invention have the following beneficial effects:

[0017] (1) Accurately assess the matching difficulty of job description-candidate resume pairs

[0018] By employing a multi-dimensional difficulty assessment mechanism, the system analyzes the difficulty of matching job descriptions with candidate resumes from three aspects: explicit information, implicit information, and fuzzy information. This ensures the comprehensiveness and accuracy of the difficulty score and provides a scientific basis for the selection of subsequent training data.

[0019] (2) Data calibration guided by business value

[0020] The innovative approach of introducing real-world behavior from hiring decision-makers as a calibration factor directly links algorithm evaluation with business performance, ensuring that the selected high-difficulty data is more valuable for training and effectively improving the model's practicality and business relevance.

[0021] (3) Optimize training efficiency and effectiveness

[0022] By establishing the compatibility relationship between reasoning style and the comprehensive difficulty score of job description-candidate resume pairs, the most valuable source data can be selected for specific supervised fine-tuning training objectives, significantly improving the efficiency of model training and final performance, while providing differentiated data selection strategies for models with different parameter scales. Attached Figure Description

[0023] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0024] Figure 1 This is a flowchart illustrating a method for selecting supervised fine-tuning training data for a job-person matching model provided in one embodiment of this application.

[0025] Figure 2 This is a schematic diagram of the structure of a device for selecting supervised fine-tuning training data for a job-person matching model provided in one embodiment of this application;

[0026] Figure 3 This is a schematic diagram of the structure of an electronic device provided in one embodiment of this application. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] One embodiment of this application provides a method for selecting supervised fine-tuning training data for a person-job matching model. The specific implementation process of this method is described in detail below. See also... Figure 1 The method includes the following steps:

[0029] Step S101: Use an artificial intelligence model to evaluate the difficulty of matching job descriptions with candidate resumes from multiple dimensions to obtain an initial difficulty score.

[0030] Traditional methods neglect the varying matching difficulty of job description-candidate resume pairs in the training data during the supervised fine-tuning phase, leading to performance degradation in complex scenarios. This step improves model robustness by establishing a multi-dimensional evaluation system to ensure that the training data covers different levels of matching difficulty.

[0031] Specifically, different job description-candidate resume pairs exhibit inherent differences in matching difficulty, which directly impacts the training performance during subsequent model-supervised fine-tuning. By designing structured evaluation dimensions, these difficulty differences can be systematically quantified. In practice, a large language model (such as GPT-4, ChatGPT, Deepseek, etc.) is first used to perform multi-dimensional matching difficulty analysis on the input job description-candidate resume pairs, obtaining a difficulty score for each dimension. The difficulty scores for each dimension are then weighted and summed to obtain the initial difficulty score for the job description-candidate resume pair.

[0032] This step involves multiple dimensions, including the difficulty of judging explicit information, implicit information, and ambiguous information. The difficulty of judging explicit information assesses the directly comparable difficulty that can be quantified, including educational background, years of work experience, and professional requirements. The difficulty of judging implicit information requires deduction and industry knowledge to assess information, including the difficulty of skill transfer, the difficulty of associating achievements, and the difficulty of completing work experience. Skill transfer difficulty assesses the feasibility and suitability of a candidate's skills across disciplines and tools; the difficulty of associating achievements assesses the strength of the connection between the candidate's project experience and the requirements of the target position; and the difficulty of completing work experience assesses the complexity of relying on common sense and industry experience to reasonably supplement information in a candidate's resume when information is missing or ambiguous. The difficulty of judging ambiguous information assesses the basic difficulty of judgment caused by vague textual descriptions or missing information.

[0033] When weighted summing the difficulty scores for each dimension, the weight of the difficulty in judging implicit information is greater than that of judging explicit information, and the weight of the difficulty in judging explicit information is greater than that of judging fuzzy information. The preferred weight allocation is: 55% for implicit information judgment difficulty, 25% for explicit information judgment difficulty, and 20% for fuzzy information judgment difficulty. This weight allocation reflects the different importance of different information dimensions in recruitment decisions, ensuring the business rationality of the difficulty score.

[0034] Step S102: Obtain matching tags generated based on the actual behavior of recruitment decision-makers, and make preliminary matching judgments through artificial intelligence models. Based on the consistency between the preliminary matching results and the matching tags, calibrate the initial difficulty score to obtain a comprehensive difficulty score.

[0035] Because there is a certain degree of disconnect between the offline annotation of the difficulty judgment of job description-candidate resume pairs by artificial intelligence models (such as large models) and real decision-making, if training data is selected solely based on the difficulty scores labeled by the artificial intelligence models, the model after supervised fine-tuning will suffer from learning bias. Therefore, this step introduces the real behaviors of recruitment decision-makers (such as downloading resumes and marking them as unsuitable). These reflect the real matching judgments in actual business scenarios and are the most valuable supervisory signals, ensuring that data selection meets actual recruitment needs. In specific implementation, user behavior logs can be obtained from the backend database of the recruitment platform. Behaviors such as "downloading resumes" and "saving resumes" are interpreted as positive feedback operations, with the corresponding matching label "match"; behaviors such as "marking as unsuitable" are interpreted as negative feedback operations, with the corresponding matching label "not match". Simultaneously, the same or different artificial intelligence models (such as large language models) as those in step S101 are used to perform preliminary matching judgments on the same job description-candidate resume pairs to obtain preliminary matching results of "match" or "not match".

[0036] Furthermore, the initial difficulty score is calibrated. When the AI ​​model's initial judgment aligns with the actual behavior of hiring decision-makers, it indicates the judgment is relatively easy. Conversely, discrepancies suggest cognitive differences or complex situations, indicating a more challenging judgment. Therefore, if the initial matching result matches the matching label, the initial difficulty score is multiplied by a weighting coefficient less than 1 (preferably 0.8) to lower the initial difficulty score; if it doesn't match, it is multiplied by a weighting coefficient greater than 1 (preferably 1.2) to raise the initial difficulty score. This calibration mechanism ensures that cases that are truly challenging in business practice receive higher difficulty scores, thus gaining priority in subsequent training data selection. The calibrated initial difficulty score is called the comprehensive difficulty score. Using the comprehensive difficulty score to filter suitable training data during the supervised fine-tuning phase reduces the number of obviously mismatched positions recommended.

[0037] Step S103: Based on the overall difficulty score, select job positions and resumes that match the difficulty level to construct a supervised fine-tuning training set for specific reasoning styles.

[0038] "Reasoning style" refers to the differences in the level of detail, logical structure, or language style reflected in the data during the model's reasoning process. These differences may be caused by the generative model, such as by large language models with different parameter sizes (e.g., 7B vs. 32B) or different architectures. They may also be caused by differences in prompts, such as by prompts that guide the generation of different instructions. Some styles emphasize the complete presentation and comparison of evidence, while others emphasize step-by-step reasoning based on the importance weight of the evidence.

[0039] A "supervised fine-tuning training set with a specific reasoning style" refers to a pre-prepared dataset with inputs and outputs in a specific format for performing supervised fine-tuning (SFT). Its core characteristic is that it requires not only that the model provide a "match / non-match" answer, but also that the model demonstrate the entire logical process of reasoning to arrive at that answer in a fixed, readable, and specific reasoning style.

[0040] For example:

[0041] Suppose we want to construct a training set for "Style A: Emphasizing stepwise reasoning based on the importance of evidence". Each training data point in this set would take the following form:

[0042] 1. Input

[0043] [Job Description]: Seeking one Senior Java Engineer. Requirements: 5+ years of backend development experience, proficient in Spring Cloud microservice architecture, experience in high-concurrency system design, Bachelor's degree or above in Computer Science or related field.

[0044] [Resume Text]: Zhang San has 6 years of Java development experience. He led the microservice transformation of an e-commerce platform, using Spring Boot and Dubbo technologies. He is familiar with MySQL and Redis. He holds a Bachelor's degree in Information Management.

[0045] 2. Expected output (i.e., reasoning style)

[0046] [Reasoning process]:

[0047] Analysis of primary core requirements:

[0048] The job requires "proficiency in Spring Cloud microservice architecture" and "experience in designing high-concurrency systems." These are the most critical technical skills for this position.

[0049] Evidence matching assessment:

[0050] Candidates must have 6 years of experience, meeting the strict requirement of "5 years or more".

[0051] The candidate has experience leading microservices transformations, but their technology stack is Dubbo rather than Spring Cloud. Although the underlying concepts are similar, the specific technical differences represent the main risk point of this mismatch.

[0052] The candidate's resume did not explicitly mention specific projects or metrics related to "high-concurrency system design," indicating insufficient experience matching.

[0053] The major and academic qualifications meet the requirements, which is a non-critical deduction item.

[0054] Overall assessment: Since the candidate failed to provide direct and compelling evidence supporting their claims regarding the two most critical technical capabilities (Spring Cloud and high concurrency), therefore...

[0055] [Final Conclusion]: Mismatch.

[0056] As can be seen from the example above, the entire "input-output" pair constitutes one training data point in the "style A training set".

[0057] The technical principle behind this step is that different reasoning styles (i.e., the characteristics of how the model outputs the reasoning process) have a specific fit relationship with training data of different matching difficulty. By optimally matching the two, the training effect can be maximized.

[0058] The specific implementation process includes two sub-steps:

[0059] Step 1: Establish a mapping relationship between "reasoning style and difficulty level".

[0060] Keeping the pre-trained model, model hyperparameters, and test set unchanged, multiple training sets are constructed using job description-candidate resume pairs with different difficulty levels for the same reasoning style. The models are then fine-tuned under supervision, and the performance metrics of each fine-tuned model are evaluated on a unified test set. The difficulty level of the training set corresponding to the model with the best performance metrics is determined as the difficulty level that is suitable for the reasoning style.

[0061] Step 2: Selecting Training Data

[0062] Based on the specific reasoning style used in the current training task, job description-candidate resume pairs with overall difficulty scores falling within the corresponding difficulty level range are selected to construct a supervised fine-tuning training set.

[0063] Furthermore, the method in this embodiment also includes:

[0064] Step S104: Based on the comprehensive difficulty score, select job description-candidate resume pairs with different appropriate difficulty levels to construct training sets for the models to be trained with different parameter scales.

[0065] Specifically, matching job descriptions to candidate resumes at different difficulty levels allows for the matching of models with different parameter sizes. This is because models with different parameter sizes have varying learning capabilities and capacity limitations. Small-parameter models have limited resources and require simple data for rapid convergence, while large-parameter models are more capable and can learn complex patterns. Therefore, datasets with higher difficulty and more complex content may be better suited to models with larger parameter sizes (e.g., 3B) due to their stronger learning and representation capabilities; conversely, datasets with lower difficulty and clearer patterns may be better suited to models with smaller parameter sizes (e.g., 0.5B), helping them achieve rapid convergence and effective learning under resource-constrained conditions. This provides a basis for differentiated data selection based on different computing resources and performance requirements in practical applications.

[0066] In practice, based on the overall difficulty score, different levels of job descriptions and candidate resumes can be selected to construct supervised fine-tuning training sets for models with different parameter sizes. For models with smaller parameter sizes (e.g., 0.5B), data with lower overall difficulty scores is prioritized; for models with larger parameter sizes (e.g., 3B), data with higher overall difficulty scores is prioritized. This allows the accuracy of the 0.5B model with small parameter sizes to approach that of larger models on suitable, simple data, while the accuracy of the 3B model with large parameter sizes is greatly improved on suitable, highly difficult data, fully leveraging the performance potential of larger models.

[0067] The following describes the complete implementation process of this embodiment in detail, using a real-world application of a large recruitment platform as an example:

[0068] (1) Data preparation stage

[0069] We extracted 100,000 job description-candidate resume pairs from the platform's historical logs, including combinations of various industries, job levels, and experience levels.

[0070] (2) Difficulty assessment stage

[0071] Using a large language model combined with structured prompts, the matching difficulty of each job description-candidate resume pair is assessed across three dimensions.

[0072] For example, when comparing the resume of a senior Java engineer with that of a candidate with 5 years of experience, the difficulty score for judging explicit information is 6.5 (the years of work experience match but the major is slightly different), the difficulty score for judging implicit information is 8.2 (a deep analysis of the transferability of the technology stack and the relevance of project results is required), and the difficulty score for judging fuzzy information is 3.1 (the information is basically complete). After weighted summation, the initial difficulty score is 7.1.

[0073] (3) Business calibration phase

[0074] The actual feedback on the platform regarding the job description and candidate resumes was that the company's HR downloaded the resumes (the matching tag was "matched"), while the initial judgment result of the large language model was "not matched". Because the initial matching result is inconsistent with the matching tag, the initial difficulty score is weighted up, that is: initial difficulty score 7.1 × weight coefficient 1.2 = overall difficulty score 8.5, which belongs to the "difficult to match judgment" level.

[0075] (4) Training set construction stage

[0076] Assume the current training task adopts "Style C: Step-by-Step Core Responsibility Matching Analysis," and that this style is known to be best suited for data at the "Match Judgment Difficulty" level through preliminary experiments. Therefore, all job description-candidate resume pairs with an overall difficulty score in the range of 7.5-10.0 are selected and used specifically to generate supervised fine-tuning training data for "Style C."

[0077] This embodiment uses a multi-dimensional difficulty assessment system to accurately quantify the complexity of matching job descriptions with candidate resumes. It introduces a business feedback closed-loop mechanism to dynamically calibrate the algorithm evaluation with real recruitment decisions, ensuring the business value orientation of training data selection. It establishes the adaptability relationship between reasoning style and the difficulty of matching job descriptions with candidate resumes, optimizes data allocation for different training objectives, and implements differentiated data selection strategies based on the characteristics of model parameter scales. This fully leverages the learning potential of models with different parameter scales, ultimately achieving a comprehensive improvement in the accuracy, generalization ability, and practicality of the job matching model, and significantly optimizing training efficiency and resource utilization.

[0078] See Figure 2In another embodiment of the present invention, a supervised fine-tuning training data selection device 200 for a job-person matching model is provided, including a model scoring module 201, a scoring calibration module 202, and a training data filtering module 203. The supervised fine-tuning training data selection device 200 for the job-person matching model is capable of executing the supervised fine-tuning training data selection method for the job-person matching model in the method embodiment.

[0079] Specifically, the device 200 for selecting supervised fine-tuning training data for the person-job matching model includes:

[0080] The model scoring module 201 is used to evaluate the difficulty of matching the job description-candidate resume pair through an artificial intelligence model in multiple dimensions, obtain the difficulty score under each dimension, and perform a weighted summation of the difficulty scores under each dimension to obtain the initial difficulty score of the job description-candidate resume pair.

[0081] The scoring calibration module 202 is used to obtain matching tags generated based on the actual behavior of the recruitment decision-maker, including positive feedback operations and negative feedback operations on candidate resumes; to perform preliminary matching judgment on the job description-candidate resume pair through an artificial intelligence model to obtain a preliminary matching result; if the preliminary matching result is consistent with the matching tag, the initial difficulty score is downweighted; if the preliminary matching result is inconsistent with the matching tag, the initial difficulty score is upweighted to obtain a comprehensive difficulty score for the job description-candidate resume pair.

[0082] The training data filtering module 203 is used to select job descriptions and candidate resumes that match the difficulty level based on the comprehensive difficulty score to construct a supervised fine-tuning training set for a specific reasoning style; wherein, the reasoning style is supervised fine-tuning training data that is differentiated in terms of the level of detail of the reasoning process, logical structure, or language style.

[0083] It should be noted that the selection device 200 for supervised fine-tuning training data of the job matching model provided in this embodiment corresponds to the technical solution that can be used to execute various method embodiments. Its implementation principle and technical effect are similar to the method, and will not be repeated here.

[0084] See Figure 3 Another embodiment of the present invention provides a schematic diagram of an electronic device 300, which is used to implement the method for selecting supervised fine-tuning training data of the person-job matching model in the method embodiment. The electronic device 300 in the embodiments of the present invention may include, but is not limited to, a PC, a server, a laptop computer, and a smart terminal. Figure 3 The electronic device 300 shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0085] like Figure 3 As shown, the electronic device 300 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes to implement the methods of the embodiments described herein, based on a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. The RAM 303 also stores various programs and data required for the operation of the electronic device 300. The processing device 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0086] Typically, the following devices can be connected to I / O interface 305: input devices 306 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 307 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 308 including, for example, magnetic tapes, hard disks, etc.; and communication devices 309. Communication device 309 allows electronic device 300 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 3 An electronic device 300 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively.

[0087] The above description is merely a preferred embodiment of the present invention. Those skilled in the art should understand that the scope of disclosure involved in the present invention is not limited to the technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalent features without departing from the above-described concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the present invention.

Claims

1. A method for selecting supervised fine-tuning training data of a human position matching model, characterized in that, The method comprises the following steps: Multi-dimensional evaluation of the matching difficulty of the position description-candidate resume pair by an artificial intelligence model to obtain difficulty scores in each dimension, and weighted summation of the difficulty scores in each dimension to obtain an initial difficulty score of the position description-candidate resume pair; Obtaining a matching label generated based on the real behavior of a recruitment decision maker, wherein the real behavior of the recruitment decision maker includes positive feedback operations and negative feedback operations on the candidate resume; Performing a preliminary matching judgment on the position description-candidate resume pair by the artificial intelligence model to obtain a preliminary matching result, and performing down-weighting processing on the initial difficulty score if the preliminary matching result is consistent with the matching label, or performing up-weighting processing on the initial difficulty score if the preliminary matching result is inconsistent with the matching label, to obtain a comprehensive difficulty score of the position description-candidate resume pair; Keeping the pre-trained model, model hyperparameters and test set unchanged, constructing multiple training sets using position description-candidate resume pairs of different difficulty levels for the same inference style, and performing supervised fine-tuning on the model to evaluate the performance indicators of each supervised fine-tuned model on the unified test set, and determining the difficulty level of the training set corresponding to the model with the optimal performance indicator as the difficulty level suitable for the inference style; selecting the position description-candidate resume pair whose comprehensive difficulty score falls within the corresponding difficulty level interval for constructing the supervised fine-tuning training set of the specific inference style according to the specific inference style adopted by the current training task; wherein the inference style is the difference in detailed degree, logical structure or language style of the inference process.

2. The method of claim 1, wherein the method further comprises: determining a plurality of candidate training data sets; and selecting a training data set from the plurality of candidate training data sets based on a performance of the human job matching model on the training data set. The step of performing down-weighting processing on the initial difficulty score if the preliminary matching result is consistent with the matching label, or performing up-weighting processing on the initial difficulty score if the preliminary matching result is inconsistent with the matching label, comprises: If the preliminary matching result is consistent with the matching label, multiplying the initial difficulty score by a weight coefficient less than 1; if the preliminary matching result is inconsistent with the matching label, multiplying the initial difficulty score by a weight coefficient greater than 1.

3. The method of claim 1, wherein the method further comprises: determining a plurality of candidate training data sets; and selecting a training data set from the plurality of candidate training data sets based on a plurality of criteria. The dimensions include explicit information judgment difficulty, implicit information judgment difficulty and fuzzy information judgment difficulty; wherein the weight of the implicit information judgment difficulty is greater than that of the explicit information judgment difficulty, and the weight of the explicit information judgment difficulty is greater than that of the fuzzy information judgment difficulty.

4. The method for selecting supervised fine-tuning training data for a person-job matching model according to claim 3, characterized in that, The implicit information judgment difficulty includes at least one of skill transfer difficulty of the job seeker, achievement association difficulty of the job seeker and work experience completion difficulty of the job seeker.

5. The method of claim 1, wherein the method further comprises: determining a plurality of candidate training data sets; and selecting a training data set from the plurality of candidate training data sets based on a plurality of criteria. The positive feedback operation is downloading the candidate resume, and the negative feedback operation is marking the candidate resume as unsuitable.

6. The method of claim 1, wherein the method further comprises: Further comprising: Selecting position description-candidate resume pairs of different adaptive difficulty levels to construct training sets of models to be trained for different parameter scales according to the comprehensive difficulty score.

7. A device for selecting supervised fine-tuning training data for a person-job matching model, characterized in that, Comprising: The model scoring module is configured to perform multi-dimensional evaluation on the matching difficulty of the position description-candidate resume pair by using the artificial intelligence model, to obtain difficulty scores in each dimension, to perform weighted summation on the difficulty scores in each dimension, and to obtain an initial difficulty score of the position description-candidate resume pair. The scoring calibration module is configured to obtain a matching label generated based on real behaviors of a recruitment decision maker, wherein the real behaviors of the recruitment decision maker include positive feedback operations and negative feedback operations on candidate resumes. The position description-candidate resume pair is preliminarily matched by using the artificial intelligence model to obtain a preliminary matching result. If the preliminary matching result is consistent with the matching label, the initial difficulty score is down-weighted. If the preliminary matching result is inconsistent with the matching label, the initial difficulty score is up-weighted to obtain a comprehensive difficulty score of the position description-candidate resume pair. The training data screening module is configured to keep the pre-trained model, model hyperparameters, and test set unchanged, to use position description-candidate resume pairs of different difficulty levels to construct multiple training sets for the same inference style, to perform supervised fine-tuning on the model, to evaluate the performance indicators of the models after the supervised fine-tuning on a unified test set, to determine the difficulty level of the training set corresponding to the model with the optimal performance indicator as the difficulty level suitable for the inference style, and to select position description-candidate resume pairs with comprehensive difficulty scores falling within the corresponding difficulty level interval for constructing a supervised fine-tuning training set of a specific inference style according to the specific inference style used in the current training task. The inference style is a supervised fine-tuning training data with differentiated details, logical structures, or language styles in the inference process.

8. The person-post matching model's supervised fine-tuning training data selection device according to claim 7, wherein, The dimensions include explicit information judgment difficulty, implicit information judgment difficulty, and fuzzy information judgment difficulty. The weight of the implicit information judgment difficulty is greater than that of the explicit information judgment difficulty, and the weight of the explicit information judgment difficulty is greater than that of the fuzzy information judgment difficulty.

Images