What the data for this category looks like
Financial report data for the chemical fiber category comes from periodic reports of listed companies disclosed by the Shanghai and Shenzhen Stock Exchanges, industry operation data released by the China Chemical Fiber Industry Association, and publicly available raw material quotation information from upstream petrochemical enterprises. Update cycles follow fixed schedules: listed company periodic reports are updated quarterly and annually, industry data is updated monthly, and raw material quotations are updated daily. Document structures include consolidated financial statements, detailed revenue breakdowns by main business segments, and capacity and production disclosure items. Fields cover revenue, capacity, production volume, and more, with units of RMB yuan, ten thousand tons, and yuan/ton respectively. A complete annual report document for a single listed company can be dozens of pages long.
What constraints these characteristics impose on the "citation sources and traceability" link
Data sources for the chemical fiber category are scattered and have varying levels of authority. This requires the traceability link to clearly mark the data source type and update time, to avoid mixing information from different channels. A single financial report contains multiple segmented detailed data, with valid information fragments distributed dispersedly. It is necessary to accurately match corresponding financial report fragments to business questions, and avoid mixing irrelevant cross-segment content into responses. In addition, industry data has exclusive field abbreviations. Original field names must be retained during traceability to ensure the accuracy and readability of citations.
How to set the configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
recall_top_k | Top 8-12 results | The chemical fiber financial report has a large number of segmented fragments. Too many recalls will introduce irrelevant content, while too few will fail to cover all segmented data required for the question |
vector_segment_length | 600-800 characters | Chemical fiber financial reports contain long tables and detailed items. Segments that are too long will lose field associations, while segments that are too short will destroy the integrity of business logic |
source_tag_mode | Tag by data source type | There are significant differences in the authority of chemical fiber data sources. It is necessary to clearly distinguish between different types of sources such as exchange announcements and industry association data |
rerank_top_k | Top 4-6 results | Segmented financial report fragments have high similarity. Reranking can filter low-relevance content and retain the most matching traceability fragments |
enable_source_fragment | Enabled | It is necessary to restore the original text fragments returned by the knowledge base or tool calls, which meets the rigor requirements of financial report analysis |
vector_model | Calibrate based on actual testing | Different vector models have varying encoding effects on chemical fiber industry terms, and adjustments should be made based on actual recall results |
The parameter values provided on this page are conventional recommendations used as starting points for configuration. Actual values are affected by material form, data volume, and business rules. Specific issues require individual analysis, and it is recommended to test on your own samples before finalizing settings.
Three common mistakes
- Phenomenon: After calling MySQL data via Function CALL, the response only returns structured conclusions without displaying the original database text fragments. Cause: The
enable_source_fragmentconfiguration is not enabled, or the original text fragment reference parameter is not bound in the tool call logic. - Phenomenon: The recalled fragments include non-chemical fiber sector petrochemical content that does not match the question. Cause: The
vector_segment_lengthis set too long, causing cross-segment long texts to be merged into one segment, which destroys the relevance of segmented data. - Phenomenon: Traceability tags only display file names without distinguishing data source types, making it impossible to judge the authority of information. Cause: The
source_tag_modeis set to only display file names, and no markers are added based on data source types.
How to confirm the configuration is correctly set
- Initiate a test conversation containing questions about revenue of chemical fiber segmented segments, check the traceability tags below the response, and confirm that each tag marks the data source type and file name.
- Check the length of the recalled fragments, confirm that the segments do not exceed the preset
vector_segment_lengthvalue, and do not contain irrelevant segment content. - Adjust the
recall_top_kvalue, test the response completeness under different recall quantities, and confirm that all segmented data fragments required for the question are covered. - Call Function CALL to obtain chemical fiber raw material cost data in MySQL, and confirm that the original text fragments returned by the database query are attached to the response.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-14.