Same-Place Image Search

Runtime and requirements

• Agent budget: 2 hours
• Compute: 8 CPUs, 32 GiB RAM, 100 GiB storage; 1 GPU(s)

Build: 3,600 seconds. Up to 140 requests in two calls of at most 70, each allowed 2,700 seconds. Keep the supplied CLIP encoder fixed. Additional pixel-based geometric verification may inspect at most 20 candidates per request. No external APIs.

Public input and output contract

Inputs

Gallery images and records.jsonl are under /task/data/corpus/; CLIP is under /task/models/clip/. Requests contain query_id, absolute query image_path and exclude_image_ids. 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 five distinct eligible image_id/finite score pairs in nonincreasing order. Respect every exclusion and process requests independently.

The completed index is read-only during retrieval.

Evaluation

Compute binary nDCG@5 against the original Hard positives after removing excluded images. Average within Oxford and Paris, then give the two sources equal weight.

This task uses the source Hard relevance protocol with a task-specific top-five metric. Invalid or excluded returned IDs invalidate that query.

Reward mapping

The final reward uses the raw quality described above, with fixed reference quality b = 0.5364309790092764 and integer score anchor a = 15 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.