dp-zp-agent/tools/graphrag/evaluation/results/final_2026-10-24/holdout_failures.json

201 lines
5.9 KiB
JSON

{
"policy": "Recorded after the frozen TEST. No implementation or audit changes were made to improve these cases.",
"diagnosis": "Both questions concern the same pair. An additional Martin Jancura document appears in the evidence; likely overly broad fuzzy person matching. Investigate in a future version using separate development cases and a new holdout.",
"questions": [
{
"count_ok": true,
"expected_count": null,
"actual_count": null,
"complete": true,
"id": "g0065",
"split": "test",
"category": "same_academic_year_pair",
"difficulty": "hard",
"question": "V ktorom rovnakom roku je v ZP Wiki uvedená práca pri osobách Ján Malinovský a Martin Šarišský?",
"latency_seconds": 0.034872,
"error": null,
"strict": false,
"value_score": 0.666667,
"entity_score": 0.0,
"relation_score": 1.0,
"property_score": 0.0,
"relation_choice_score": 0.0,
"document_recall": 1.0,
"source_recall": 1.0,
"expected_values": [
"Ján Malinovský",
"Martin Šarišský",
"2026"
],
"matched_values": [
"Ján Malinovský",
"Martin Šarišský"
],
"expected_entities": [
{
"type": "year",
"value": "2026"
}
],
"matched_entities": [],
"expected_relations": [],
"actual_relations": [],
"matched_relations": [],
"expected_properties": [
"Thesis.year|Work.year",
"Thesis.year|Work.year"
],
"matched_properties": [],
"relation_choices": [
{
"choices": [
"AUTHORED",
"HAS_WORK"
],
"matched": false
},
{
"choices": [
"AUTHORED",
"HAS_WORK"
],
"matched": false
}
],
"expected_documents": [
"pages/students/2021/martin_sarissky/README.md",
"pages/students/2022/jan_malinovsky/README.md"
],
"actual_documents": [
"pages/students/2017/martin_jancura/README.md",
"pages/students/2021/martin_sarissky/README.md",
"pages/students/2022/jan_malinovsky/README.md"
],
"matched_documents": [
"pages/students/2021/martin_sarissky/README.md",
"pages/students/2022/jan_malinovsky/README.md"
],
"expected_source_urls": [
"https://zp.kemt.fei.tuke.sk/students/2021/martin_sarissky",
"https://zp.kemt.fei.tuke.sk/students/2022/jan_malinovsky"
],
"actual_source_urls": [
"https://zp.kemt.fei.tuke.sk/students/2017/martin_jancura",
"https://zp.kemt.fei.tuke.sk/students/2021/martin_sarissky",
"https://zp.kemt.fei.tuke.sk/students/2022/jan_malinovsky"
],
"matched_source_urls": [
"https://zp.kemt.fei.tuke.sk/students/2021/martin_sarissky",
"https://zp.kemt.fei.tuke.sk/students/2022/jan_malinovsky"
],
"entity_set_expected": [],
"entity_set_actual": [],
"entity_set_mode": null,
"entity_set_ok": true,
"strategies": [
"entity_resolution"
],
"fact_types": [
"academic_year_comparison"
]
},
{
"count_ok": true,
"expected_count": null,
"actual_count": null,
"complete": true,
"id": "g0129",
"split": "test",
"category": "same_academic_year_pair",
"difficulty": "hard",
"question": "Aký rok práce spája osoby Ján Malinovský a Martin Šarišský?",
"latency_seconds": 0.032707,
"error": null,
"strict": false,
"value_score": 0.666667,
"entity_score": 0.0,
"relation_score": 1.0,
"property_score": 0.0,
"relation_choice_score": 0.0,
"document_recall": 1.0,
"source_recall": 1.0,
"expected_values": [
"Ján Malinovský",
"Martin Šarišský",
"2026"
],
"matched_values": [
"Ján Malinovský",
"Martin Šarišský"
],
"expected_entities": [
{
"type": "year",
"value": "2026"
}
],
"matched_entities": [],
"expected_relations": [],
"actual_relations": [],
"matched_relations": [],
"expected_properties": [
"Thesis.year|Work.year",
"Thesis.year|Work.year"
],
"matched_properties": [],
"relation_choices": [
{
"choices": [
"AUTHORED",
"HAS_WORK"
],
"matched": false
},
{
"choices": [
"AUTHORED",
"HAS_WORK"
],
"matched": false
}
],
"expected_documents": [
"pages/students/2021/martin_sarissky/README.md",
"pages/students/2022/jan_malinovsky/README.md"
],
"actual_documents": [
"pages/students/2017/martin_jancura/README.md",
"pages/students/2021/martin_sarissky/README.md",
"pages/students/2022/jan_malinovsky/README.md"
],
"matched_documents": [
"pages/students/2021/martin_sarissky/README.md",
"pages/students/2022/jan_malinovsky/README.md"
],
"expected_source_urls": [
"https://zp.kemt.fei.tuke.sk/students/2021/martin_sarissky",
"https://zp.kemt.fei.tuke.sk/students/2022/jan_malinovsky"
],
"actual_source_urls": [
"https://zp.kemt.fei.tuke.sk/students/2017/martin_jancura",
"https://zp.kemt.fei.tuke.sk/students/2021/martin_sarissky",
"https://zp.kemt.fei.tuke.sk/students/2022/jan_malinovsky"
],
"matched_source_urls": [
"https://zp.kemt.fei.tuke.sk/students/2021/martin_sarissky",
"https://zp.kemt.fei.tuke.sk/students/2022/jan_malinovsky"
],
"entity_set_expected": [],
"entity_set_actual": [],
"entity_set_mode": null,
"entity_set_ok": true,
"strategies": [
"entity_resolution"
],
"fact_types": [
"academic_year_comparison"
]
}
]
}