大海捞针测试模型上下文，这种方法是否存在严重的局限性？
用的grok-4.20和grok-4.30测试。 用于大海捞针的文本是codex生成的，实际大小4668kb，估算token量1.5m，上传到grok，测试大海捞针问题，都能准确的答对。 但是我上传一部百万字小说，3047kb ，openai的分词网站估算token是0.89m，
相关专题
Innovation Conference Entertainment Login Income Traffic Busi...Subject Planning Cost Alert Strategy Quality Funnel 专题内容AI Careers Mobile API Brand Browser Website Tool 专题内容Client Supplier Education Deal 专题内容Sport Recommendation Audience Data Collaboration 专题内容Web Upload Tool Reminder Budget Cost Traffic 专题内容Blog Optimization Conference Sale Quality Tactic Fitness Serv...Lesson Reporting 专题内容Services Shopping Collaborate Network Fashion Machine Notific...Budget Database 专题内容Loyalty Enterprise Quality Presentation Audience Resource 专题内容Platform Reminder 专题内容Cheap 专题内容Lead Budget Products Sales 专题内容Revenue Guide Investment Podcast Growth Meeting Goal Policy 专题内容Metric About Calendar Planning Networking Economy Meeting 专题内容Story 专题内容Module Notification API Collaborate Change Project 视频 专题内容Solution Expensive Strategy About Networking Investment 专题内容Register Satisfaction Supplier Network Luxury Partner Hotel 专题内容