Citation and Traceability for CMC Research Registration and

Data Characteristics in this Category

CMC (Chemistry, Manufacturing, and Control) research registration and declaration documents involve data from the entire lifecycle of new or generic drugs, from R&D to manufacturing. Data sources are diverse. They include laboratory research reports, pilot production records, quality standards and inspection reports, stability study data, and process validation documents. These documents are typically in formats like PDF, Word, and Excel, with varying degrees of structure. Data updates frequently, especially during early-stage R&D and clinical trials, as process parameters, quality attributes, and stability results continuously iterate. Documents contain extensive specialized terminology, chemical structures, charts, and units (e.g., mg/mL, ℃, pH value, HPLC peak area). Accuracy and consistency requirements are extremely high.

Constraints from these Characteristics on "Citation and Traceability"

The diversity and complexity of CMC data pose challenges for citation and traceability. First, the heterogeneous document formats require the knowledge base to have robust multi-format parsing capabilities to ensure all key information is extracted effectively. Second, the highly specialized nature of the data and its unit sensitivity mean that citations must be precise down to specific values and units, avoiding semantic ambiguity. High update frequency requires the knowledge base to quickly identify and update related information, ensuring that citations always point to the latest version. Furthermore, due to the rigor of CMC documents, traceability must pinpoint the specific page number or paragraph of the original document, or even specific charts or tables, to meet regulatory requirements for data integrity and traceability.

Configuration Guidelines

Configuration Item	Recommended Value	Rationale
`Chunk size` (Segment Length)	`800–1200 characters`	CMC reports often have long paragraphs containing multiple contextual pieces of information. Semantic completeness must be maintained.
`Recall count` (Recall Count)	`Top 10`	Ensures coverage of key information from multiple relevant documents, preventing omissions.
`Similarity threshold` (Similarity Threshold)	`0.75–0.85`	Balances relevance with precise matching of specialized terminology.
`Rerank result count` (Reranked Return Count)	`Top 5`	Selects the most relevant content from the recalled results, improving the accuracy of final citations.
`PARSE_FILE_TIMEOUT_SECONDS`	`600 seconds`	Parsing large PDF files is time-consuming. Sufficient time is allocated to prevent timeout interruptions.
`UPLOAD_FILE_MAX_SIZE`	`200 MB`	Accommodates declaration documents containing numerous charts, graphs, and scanned images.

Three Common Pitfalls

Citation results contain a large amount of irrelevant or duplicate content. This occurs when the Similarity threshold (Similarity Threshold) is set too low, leading to the recall of semantically imprecise document segments.
Some nodes in the workflow fail to retrieve knowledge base citations, indicated by missing citation information in the returned content. This may be due to issues with the preceding node's datasetid variable passing, preventing the knowledge base retrieval node from correctly identifying the target dataset.
Critical numerical values or units are missing from citation results. This happens when the original document parsing quality is poor, failing to accurately extract specific table or chart data containing numbers and units.

How to Verify Correct Configuration

Test with a CMC report containing complex charts and specialized terminology. Verify that citation results accurately pinpoint chart titles or table data.
Upload a revised version of the same report. Confirm that after the update, citation results point to the latest content and differentiate between new and old versions.
Simulate questions that require information from multiple associated documents for a complete answer. Check if citation results can extract relevant paragraphs from different documents simultaneously and trace them back to their respective original files and page numbers.
Query text containing specific chemical structures or formulas. Verify that citation results accurately display these special characters and link back to the original source.

Note: The values provided are common starting points. They should be measured against the reader's own samples.

Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-21.