Narrative-Clue Search

Runtime and requirements

• Agent budget: 2 hours
• Compute: 8 CPUs, 32 GiB RAM, 100 GiB storage; 0 GPU(s)

Build: 3,600 seconds. Up to 450 requests in two calls of at most 225; each call has 2,700 seconds. Only the documented OpenRouter/Jina APIs may supplement local computation.

Public input and output contract

Inputs

/task/data/corpus/records.jsonl contains doc_id, title and article text. Queries contain query_id and question. Public sentence annotations also appear in /task/data/training/sentence_labels.jsonl. Public requests and labels are under /task/data/validation/.

Outputs

Submit executable /app/build.sh and /app/run.sh and all supporting files under /app. The fresh verifier rebuilds the index and runs commands as a non-root user with code and inputs read-only. Honor the supplied index, output, TMPDIR and cache paths; the index parent is not writable.

/app/build.sh --corpus /task/data/corpus --index-dir /work/index
/app/run.sh --index-dir /work/index --queries /path/to/queries.jsonl --output /path/to/results.jsonl

Return query_id and at most ten distinct doc_id/finite score results, ordered by nonincreasing score.

The index remains writable across repeated query calls.

Evaluation

Compute nDCG@10 using complete original qrels, then average equally over 445 held-out questions.

Four category means based on original human uncertainty annotations and a fixed median narrative length are diagnostics. They do not change query weights or introduce model-generated categories.

Reward mapping

The final reward uses the raw quality described above, with fixed reference quality b = 0.40178618363049623 and integer score anchor a = 20 on the 0–100 scale:

reward(q) = (a / 100) * q / b                             if 0 <= q <= b
            a / 100 + (1 - a / 100) * (q - b) / (1 - b)  if b < q <= 1

This maps raw quality 0, b and 1 to rewards 0, a / 100 and 1. Raw quality and auxiliary diagnostics are reported separately. Harbor retains the 0–1 reward precision; a displayed 0–100 score is rounded to one decimal place. Correctness and trajectory-audit gates take precedence over this conversion.

Validity and failure handling

The verifier checks successful index construction, executable behavior, query coverage, IDs, duplicate handling, finite scores and stage limits. Completed valid JSONL rows remain usable after a query timeout; missing or invalid rows score zero. A non-timeout query-command failure invalidates that call. See instruction.md and the verifier for the exact task-specific validity checks.

Final reward requires a healthy, accepting independent trajectory and submission integrity audit. Audit failures are reported separately from retrieval quality and cannot produce positive credit.

See the full instructions, task documentation and runtime configuration for the complete contract at the synchronized benchmark revision.