README.md

metadata

license: mit
configs:
  - config_name: test_queries
    default: true
    data_files:
      - split: test
        path: test.jsonl

DAPS-Blueprint

• full_collection/ contains the complete Blueprint node records and their dependencies, organized by source project. These files provide the graph used for training and premise retrieval.
• splits.json records the fixed random split of 2,351 query nodes at the node level. Its recorded seed is 42, with an 80% / 10% / 10% train / validation / test ratio: 1,880 training nodes, 235 validation nodes, and 236 test nodes. Use this file to select the nodes for each split.
• test.jsonl exports those 236 test queries and their gold positive premise nodes for inspection and independent evaluation. Its query order matches the test split in splits.json.

The complete node collection is retained for graph context and premise retrieval; the split file identifies the query nodes used in each stage.