What this category's data looks like
This page addresses the enterprise new-employee onboarding training scenario, and the materials include three types: onboarding training plans, course sign-in sheets, and job descriptions. The onboarding training plan gives training topics, class-hour arrangements, and responsible persons; the course sign-in sheet records participants, dates, and signatures; the job description lists job responsibilities, qualification requirements, and the effective version. The three types of materials are updated at different rhythms: training plans are adjusted by batch, sign-in sheets are generated by session, and job descriptions are revised as the organization changes. When the materials enter the business basis query function, the training plan plays the role of the basis for training arrangements, the sign-in sheet plays the role of the basis for attendance and acknowledgments, and the job description plays the role of the basis for job responsibilities and versions. Field characteristics are concentrated in four categories: names, quantities, dates, and signatures. When comparing across materials, attention must be paid to whether the same record is expressed consistently in the three materials, whether the version is current, and whether signatures or acknowledgments are complete. Query results need to form basis entries that can point back to the original text, for use in replying to parents or writing into proposals.
Configuration and implementation checks
This link is executed by application settings (response limit, etc.), which output the length of a single answer and the range of context that can be accommodated. "Basis entries that can point back to the original text" are then completed by the model. The response limit is controlled by a slider for the length of a single answer; if set too low, the answer will be truncated, and if set too high, consumption will increase.
| Object | Check | Verification action |
|---|---|---|
| Response limit | Matches the answer length required by the page | In application settings, check the switch and slider position, and confirm that the single-answer length and context range meet the query needs |
| Truncation behavior | The ending of a longer answer is complete | Ask a cross-material basis question that requires a long answer, see whether the ending is complete, and check the values of maxResponse and maxContext |
| Runtime consumption | The single-run usage changes after the limit is adjusted | Compare the usage of two runs, record the quoteMaxToken budget and the number of quoted entries, and record the number of array entries separately |
| Onboarding training plan and course sign-in sheet | Names, quantities, dates, and signatures can be matched to the original text | Compare training topics, number of participants, dates, and signatures one by one, and confirm that the same record is consistent in the two materials |
Three common errors
- Phenomenon: The name, quantity, or date of the same record does not match between the training plan and the sign-in sheet. Cause: The two materials are maintained separately, and no cross-material comparison was performed.
- Phenomenon: The job description version is not synchronized, and an old version of responsibilities is cited. Cause: The effective version marker was not updated after revision.
- Phenomenon: The sign-in sheet lacks a signature or acknowledgment, and the basis entry cannot point back to the original text. Cause: Signatures were not added or archived in time after the session ended.
How to verify the setup
This function's acceptance: Ask a question that requires a long answer, see whether the ending is complete, confirm that the response limit and context range match, and check maxResponse and maxContext.
Business test: Ask a cross-material basis question, see whether the answer gives a source that can be pointed back to, and confirm that the number of quoted entries and the quoteMaxToken budget are recorded separately.
- In application settings, check the response limit switch and slider position, and confirm that the single-answer length and context range meet the query needs.
- Compare the usage of two runs, record the number of quoted entries and the
quoteMaxTokenbudget, and confirm that truncation behavior and runtime consumption meet expectations.