Timeline Eval prompt

You are generating a timeline for evaluation.

Query title: "{QUERY_TITLE}"
Prompt profile: decompose_then_strict_selfcheck_no_k

Input setting:
- Query-only generation (no external context is provided).
- Use only the query title and your internal knowledge.

Output format (TL17, strict):
1) Each entry starts with exactly one line: YYYY-MM-DD
2) Then write 1 to 2 summary lines for that date
3) Insert exactly one blank line between entries
4) Do not output any text before, after, or between entries

Rules:
1) Work internally in two passes:
   - Pass 1: list key entities and candidate dated milestones.
   - Pass 2: keep only milestones that pass a strict check (date confidence high, directly relevant, non-duplicate, chronology-consistent).
2) Do not reveal internal passes or reasoning.
3) Use only high-confidence, well-established facts relevant to the query title.
4) If confidence in exact date is low, drop the event.
5) Keep dates unique and strictly ascending.
6) Prefer institutional/legal/official actions over interpretive commentary.
7) Use concise factual wording; avoid speculative motives and unverifiable details.
8) Entry-count policy: target 10-16 entries for balanced precision and coverage.
9) Hard cap: never output more than 24 date entries.
10) Stop immediately after the last valid entry.
11) Never repeat a date line. If multiple events happen on the same day, merge them into one date block.