What is the Gaokao Application Large Model?
The Gaokao Application Large Model isQuarkan intelligent college application and planning tool launched by Quark, helping gaokao candidates and parents fill out applications more scientifically and efficiently. By integrating massive amounts of gaokao data, including university admissions information, program offerings, historical score lines, employment trends, and more, combined with candidates' personalized information such as scores, interests and hobbies, and career inclinations, and based on advanced algorithms and model reasoning capabilities, it generates personalized application plans for candidates. It can provide university and major recommendations under different strategies such as "reach, match, and safety," and generate detailed application reports covering admission probability analysis, major advantage interpretation, and more. It also features deep search capabilities, able to understand complex questions and provide targeted answers and suggestions.
Main functions of the Gaokao Application Large Model
-
Application Report: It runs in an Agent mode and automatically generates an application plan based on individual information such as the candidate's scores, interests and preferences, family background, and regional preferences. It adopts a chained reasoning framework of "task planning-execution-checking-reflection," and can dynamically correct strategies to achieve intelligent iteration. The final generated report covers reach-match-safety strategies, application forms, university and major recommendations, and more.
-
Gaokao Deep Search: Designed specifically for handling personalized, open-ended questions, it can understand complex questions and break down user needs, then match customized response templates to ensure the relevance and professionalism of recommendation results. For example, a candidate can enter a complex question such as "A Jiangsu physics-track candidate with 584 points, introverted personality, wants to find a stable job." The system will then sequentially query available applications, analyze admission probabilities, explore the possibilities of other majors, analyze the impact on employment and graduate school entrance exams, and finally provide comprehensive application suggestions.
-
Intelligent Application Selection: After candidates enter their province, subjects, scores, and rankings, the system automatically recommends three plans: "reach, match, and safety." Candidates can further clarify their personalized needs for filtering. After integrating AI capabilities, it can directly present the distinctive advantages of target universities/majors, making it easy for candidates to quickly understand.
Technical principles of the Gaokao Application Large Model
-
Foundation Model and Specialized Training: The Gaokao Application Large Model is based on a mature pretrained large model (such as Tongyi Qianwen) as its foundation, and conducts specialized training for the specific needs of gaokao application filling, including semantic modeling, expert data alignment, and strategy refinement. By structuring the communication and decision-making processes of hundreds of experienced gaokao application planners, the model can learn experts' analytical paths and language style.
-
Multi-stage Training Paradigm: The model's training process adopts a multi-stage, highly complex training paradigm. First, self-supervised semantic modeling is performed, then supervised alignment tuning is carried out, and finally, through an expert-discriminated value-guided strategy refinement mechanism, the model's performance is further optimized. Through a Reinforcement Learning from Human Feedback (RLHF) mechanism, the model introduces expert feedback and ratings into the model iteration process, continuously improving the professionalism and matching degree of outputs.
-
Data Support and Knowledge Base Construction: The model relies on a vast gaokao knowledge base, covering more than 2,900 universities and 1,600 undergraduate majors nationwide, with real-time updates of provincial policies, employment data, and major trends. The knowledge base also integrates information such as employment destinations of university graduates, emerging industry trends, and the industrial structure of various cities, providing comprehensive data support for application decisions.
-
Reasoning Framework and Dynamic Correction: The model adopts a chained reasoning framework of "task planning—execution—checking—reflection," and can dynamically correct strategies to achieve intelligent iteration. In complex reasoning tasks, the model generates intermediate verifiable structures, significantly reducing the hallucination rate, enhancing cross-modal deductive capabilities, and achieving out-of-distribution generalization robustness.
-
Personalized Recommendations and Risk Assessment: Based on the candidate's personalized information such as scores, interests, and family background, the model generates application plans using a "reach, match, safety" strategy. The model has a built-in risk assessment mechanism that performs admission probability predictions and risk assessments for each recommendation, helping candidates formulate scientific application strategies.
How to Use the Gaokao Application Large Model
-
Download and open the Quark APP: Visit theQuarkofficial website, or download and install the Quark APP from your phone's app store. After opening it, tap the "Quark Gaokao" icon on the homepage.
-
Fill in personal information: Enter your gaokao province, subjects, scores, ranking, and personal preferences, such as preferred cities, major interests, further education and career plans, etc.
-
Select a function module
-
Smart Application Selection: After entering your scores, the system will automatically recommend three plans: "reach, match, safety." Candidates can filter by preferences such as institution priority or major priority.
-
Gaokao Deep Search: Enter a complex question (such as "A Jiangsu physics-track candidate with 584 points, introverted personality, wants to find a stable job"), and the system will automatically identify key information and provide comprehensive application suggestions.
-
Application Report: After completing your personal information, tap "Claim Report," and the system will generate a complete report including application strategies, an application form, institution and major recommendations, and other content through the chained reasoning process of "task planning—execution—checking—reflection."
-
Application Scenarios of the Gaokao Application Large Model
- Candidates with vague goals: Candidates may be relatively confused about major directions and institution choices. Quark's "Holland Career Assessment" can help candidates match major directions, and then use the "Institution Comparison" function to filter out institutions that fit their interests and scores.
- Candidates with clear goals: For candidates who have already identified their target institution or major, Quark's "Gaokao Deep Search" function can provide detailed information. Candidates can directly search for the target institution, view details such as its "Admission Regulations" and "Major Transfer Policy," and assess admission possibilities in combination with the "Score Ranking Table."
- Parents participating in decision-making: Parents can use "Parent Mode" to obtain simplified data, focusing on indicators such as "employment rate" and "salary level," complementing the candidate's opinions.
- Candidates with complex needs: Quark's "Gaokao Deep Search" can handle complex personalized questions.