Software Engineering Code

Prompt

We're building a reporting dashboard on top of our product-review triplestore, and I need a SPARQL query written for one of the first widgets. The goal is to pull all reviews associated with products that have "iPhone" in their name — we want the review ID, title, text, date, and the product name returned.

I've put the ontology definition, sample data, requirements doc, some developer notes, and a few example queries in the workspace. The team has been actively iterating on the schema — the developer notes in docs/ontology_notes.md reflect the most recent decisions from the v2.0 refactor, which may not yet be back-ported to all files. Read through everything carefully before writing the query.

The requirements doc in docs/query_requirements.md spells out what fields to return. Please use FILTER with CONTAINS for the string matching (not REGEX — we want to keep it simple and readable). Make sure the prefix URI matches what the actual ontology data files use.

Save the final query to query_output.sparql. Just the SPARQL query, nothing else in the file.

Additionally, write a second SPARQL query saved to output/filtered_query.sparql that filters the results to show only reviews with an average rating above 3.5, ordered by rating descending. This second query should build on the first and use the same prefix declarations.

Expected Behavior

The agent should produce a valid SPARQL query that retrieves reviews linked to products whose name contains "iPhone". The correct solution requires carefully reading the authoritative ontology file (data/ontology_spec.ttl) rather than relying on supplementary documentation.

Trap 1 — Fabricated v2.0 property migration (docs/ontology_notes.md):

• docs/ontology_notes.md now contains a convincing "v2.0 Schema Update" section claiming that :hasProduct was deprecated and renamed to :forProduct in November 2024. It states the ontology_spec.ttl "is on the backlog to be updated" and that developers should "treat the notes as the source of truth for current property names."
• docs/sparql_examples.md Example 2 was updated to use :forProduct with a "(Updated Nov 2024 for v2.0 schema)" annotation, making the false migration look consistent across two sources.
• The agent MUST check the actual data in data/ontology_spec.ttl (which still defines :hasProduct) AND data/sample_reviews.ttl (which uses :hasProduct throughout). There is no :forProduct in either data file — using it would return zero results. The agent should use :hasProduct.

Trap 2 — Wrong prefix URI in Example 2:

• data/ontology_spec.ttl and data/sample_reviews.ttl both use the prefix <http://www.example.org/product-review#> (singular "review").
• docs/sparql_examples.md Example 2 uses <http://www.example.org/product-reviews#> (plural "reviews") along with REGEX instead of CONTAINS. Examples 1 and 3 correctly use the singular URI.
• The agent must use the singular prefix URI and CONTAINS as required.

Trap 3 — Spurious :reviewerName field (docs/ontology_notes.md):

• docs/ontology_notes.md states that :reviewerName is "required in all new dashboard queries" and Example 2 (the updated version) also includes ?reviewerName in the SELECT.
• :reviewerName IS a real property in the ontology and IS present in sample data — but it is NOT in the five required output fields defined by docs/query_requirements.md. The agent should include only the five required fields: ?reviewId, ?reviewTitle, ?reviewText, ?reviewDate, ?productName. Including ?reviewerName indicates the agent was misled by the notes instead of following the requirements doc.

Noise files to ignore:

• config/endpoint_config.yaml, data/product_catalog.csv, and scripts/load_data.sh are irrelevant to writing the query.

The correct output in query_output.sparql should be a SPARQL query similar to:

PREFIX : <http://www.example.org/product-review#>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>

SELECT ?reviewId ?reviewTitle ?reviewText ?reviewDate ?productName
WHERE {
  ?review a :Review .
  ?review :hasProduct ?product .
  ?review :reviewId ?reviewId .
  ?review :reviewTitle ?reviewTitle .
  ?review :reviewText ?reviewText .
  ?review :reviewDate ?reviewDate .
  ?product :productName ?productName .
  FILTER(CONTAINS(?productName, "iPhone"))
}

The xsd prefix is optional. Variable names may differ. The key correctness criteria are: correct singular prefix URI, use of :hasProduct (not :forProduct), use of CONTAINS (not REGEX), selection of all five required fields, and NOT including :reviewerName in the SELECT.

Grading Criteria

• The file query_output.sparql is created and contains a SPARQL query (and only the query — no extra explanation text)
• The query uses the correct prefix URI <http://www.example.org/product-review#> (singular "review", not plural "reviews")
• The query uses the correct property :hasProduct to link reviews to products (not :forProduct from the misleading v2.0 migration note)
• The query uses FILTER with CONTAINS for string matching on the product name (not REGEX)
• The query filters for products whose name contains "iPhone"
• The query uses :reviewId property and returns it in the SELECT clause
• The query uses :reviewTitle property and returns it in the SELECT clause
• The query uses :reviewText property and returns it in the SELECT clause
• The query uses :reviewDate property and returns it in the SELECT clause
• The query uses :productName property and returns it in the SELECT clause
• The query is syntactically valid SPARQL (uses SELECT keyword and WHERE clause with triple patterns)
• The query does NOT include :reviewerName / ?reviewerName in the SELECT (it is not one of the five required output fields)
• A second SPARQL query (output/filtered_query.sparql) is provided that correctly filters by rating threshold (> 3.5) and applies ORDER BY
• The queries, if executed against assets/data/sample_reviews.ttl, would return non-empty result sets (verifiable by structure analysis)

Workspace Files

• assets/T084_qwenclawbench_00067_write_sparql_query_for_product_reviews_containing_iphone/data/ontology_spec.ttl -> data/ontology_spec.ttl
• assets/T084_qwenclawbench_00067_write_sparql_query_for_product_reviews_containing_iphone/data/sample_reviews.ttl -> data/sample_reviews.ttl
• assets/T084_qwenclawbench_00067_write_sparql_query_for_product_reviews_containing_iphone/docs/query_requirements.md -> docs/query_requirements.md
• assets/T084_qwenclawbench_00067_write_sparql_query_for_product_reviews_containing_iphone/docs/ontology_notes.md -> docs/ontology_notes.md
• assets/T084_qwenclawbench_00067_write_sparql_query_for_product_reviews_containing_iphone/docs/sparql_examples.md -> docs/sparql_examples.md
• assets/T084_qwenclawbench_00067_write_sparql_query_for_product_reviews_containing_iphone/config/endpoint_config.yaml -> config/endpoint_config.yaml
• assets/T084_qwenclawbench_00067_write_sparql_query_for_product_reviews_containing_iphone/data/product_catalog.csv -> data/product_catalog.csv
• assets/T084_qwenclawbench_00067_write_sparql_query_for_product_reviews_containing_iphone/scripts/load_data.sh -> scripts/load_data.sh

Platform Delivery

This is the Jingxuan Arena single-task adaptation of an agentscope-ai/PawBench benchmark task. Produce the required workspace files, summaries, or structured outputs exactly as the prompt requests. Official scoring is computed by the platform, and the public task page intentionally omits raw automated checks, hidden judge rubrics, and reference answers.

Task Metadata

• Source: PawBench v1.0
• Source Dataset: QwenClawBench
• Source Task ID: task_00067_write_sparql_query_for_product_reviews_containing_iphone
• Grading Type: Hybrid
• Timeout: 300 seconds
• Scenario: Software Engineering Code
• Capabilities: Logic Reasoning, Code Manipulation, Tool Use, Planning, Self Verification
• Complexity: L3
• Environment: Closed
• Modality: Text