I work part time in Hong KongJobs and Career
How each model performed on this task. Every chart states its N, and the reward carries a 95% confidence interval once there is more than one run to measure.
Core metrics
Check Pass %
Mean share of the task's checks passing, averaged over runs. N = 1 run per model.
83%
Claude Opus 5.5
67%
Claude Opus 5
50%
GPT-6 sol
42%
GPT-5.6 sol
Extended metrics
Median Time
Wall-clock minutes, median across runs. N = 1 run per model.
20m
Claude Opus 5
14m
GPT-5.6 sol
Tool Calls
Tool invocations, median across runs — counted from the trajectories. N = 1 run per model.
94
Claude Opus 5
45
Claude Opus 5.5
34
GPT-6 sol
34
GPT-5.6 sol
Verifier × model pass rate
How often each verifier passed, per model. Green is passing, red is failing — a red band across a row is a verifier nothing clears.
| Verifier | Claude Opus 5.5n=1 | GPT-6 soln=1 | GPT-5.6 soln=1 | Claude Opus 5n=1 |
|---|---|---|---|---|
| checks (12) | ||||
| 100% | 100% | 100% | 100% | |
How to passOpen https://www.labour.gov.hk/eng/news/EAO2025.htm and record the previous continuous contract test in full, the revised test with both of its alternative working hours limbs, the date the amendment was passed, the date the new requirement applies from, and who carries the onus of proof in a dispute about whether a contract is continuous. | ||||
| 100% | 0% | 100% | 100% | |
How to passOpen https://www.labour.gov.hk/eng/faq/cap57b_whole.htm and compare its statement of the continuous contract test with the notice from step 1. Record the two footnotes it attaches, including the condition about the four-week period, and state whether the two pages agree. Note anything the FAQ adds that the notice does not say. | ||||
| 100% | 0% | 0% | 0% | |
How to passTake a worker employed by the same employer throughout who worked 18 hours in each of the three immediately preceding weeks, then 16, 20, 16 and 18 hours across four consecutive weeks. Determine whether that pattern satisfies the old test and whether it satisfies the new test, applying each limb explicitly and showing the four-week total. State which weeks trigger the second limb and why the two tests reach the results they do. | ||||
| 100% | 100% | 100% | 100% | |
How to passFrom the FAQ page in step 2, record which rights the ordinance gives to every covered employee regardless of job title or working hours, and which further rights arise only under a continuous contract. Record what the page says about temporary, part-time and substituted employees, and name at least three entitlements in each group. | ||||
| 100% | 100% | 0% | 100% | |
How to passOpen https://www.labour.gov.hk/eng/faq/cap57i_whole.htm and record the full years-of-service to annual-leave-days table, from the first year through the maximum. State the service length at which the entitlement stops increasing, and state the entitlement for a worker in their fourth year and in their ninth year. | ||||
| 100% | 0% | 0% | 0% | |
How to passFrom the same page, record who appoints the timing, the notice period required and whether it can be shortened, the two splitting rules and the leave-entitlement threshold that separates them, and the only portion of the entitlement an employee may lawfully accept payment for instead of taking. Apply that last rule to a worker entitled to twelve days. | ||||
| 100% | 0% | 0% | 0% | |
How to passFrom the same page, record the basis of the daily rate, the reference period used, what happens when the employee has been employed for less than that period, the two categories that must be excluded when averaging, and by when the payment must be made. Name at least four of the excluded periods. | ||||
| 100% | 100% | 0% | 100% | |
How to passFrom the same page, record the service condition that triggers pro-rata annual leave pay when employment ends, and the one form of dismissal that removes the entitlement. State whether resignation and dismissal are treated the same for this purpose. | ||||
| 0% | 0% | 0% | 100% | |
How to passOpen https://www.labour.gov.hk/eng/labour/Statutory_Employment_Entitlements_Reference_Calculator.htm , accept the notice, answer the basic information questions for someone still employed under a continuous contract who started roughly three years before today, and continue. Record which entitlement calculators are offered, the termination entitlement table displayed for annual leave, and the sentence limiting how the output may be used. | ||||
| 100% | 100% | 100% | 0% | |
How to passReturn to the FAQ page from step 2 and record what happens to a contract term that reduces a right conferred by the ordinance. Explain the rule using the official wording and identify any qualification or exception stated on the page. | ||||
| 0% | 100% | 100% | 100% | |
How to passReturn to the FAQ page from step 2 and record what happens to a contract term that reduces a right conferred by the ordinance, the employer's obligation regarding a copy of a written contract and the fine for failing it, and the factors used to distinguish an employee from a self-employed contractor. State who has the final say when that distinction is disputed. | ||||
| 100% | 0% | 0% | 100% | |
How to passAssemble a document with one table of entitlements, marking for each whether it needs a continuous contract, the qualifying condition, and the page it came from. Above the table, state the old and new tests side by side with the changeover date and the result for the pattern in step 3. Close with the two questions the worker should put to the employer in writing, each tied to a specific rule. | ||||
Failure modes on this task
19 tagged verifier failures, by the stage they were assigned to. Higher is worse.
| Failure stage | Claude Opus 5.5 | GPT-6 sol | GPT-5.6 sol | Claude Opus 5 |
|---|---|---|---|---|
| Standing up the evidence1 mode | — | 2 | 4 | 3 |
| Delivering the result1 mode | — | 3 | 1 | 1 |
| No signal in the judge’s words1 mode · the grader | 2 | 1 | 2 | — |