多模型作文批改
教育多個 AI 模型各自對作文進行評分和評語，由仲裁模型綜合各家意見產出最終成績和改善建議。
agentsystem
為什麼需要 OSOP
單一模型的評分容易有偏差。OSOP 定義多模型批改的協作架構和仲裁機制，讓評分過程透明公平，並記錄每個模型的評分細項供教師參考。
Workflow Steps (7)
1
Essay Intake
system2
Grader A (GPT-4o)
agent3
Grader B (Claude)
agent4
Grader C (Gemini)
agent5
Score Aggregation
system6
Bias Detection Agent
agent7
Final Grade & Feedback
agentConnections (9)
Essay Intake→Grader A (GPT-4o)parallel
Essay Intake→Grader B (Claude)parallel
Essay Intake→Grader C (Gemini)parallel
Grader A (GPT-4o)→Score Aggregationparallel
Grader B (Claude)→Score Aggregationparallel
Grader C (Gemini)→Score Aggregationparallel
Score Aggregation→Bias Detection Agentsequential
Bias Detection Agent→Final Grade & Feedbackconditionalbias.detected == false
Bias Detection Agent→Grader A (GPT-4o)fallbackBias detected, re-grade with adjusted prompts
7
Steps
9
Connections
2
Node Types