File size: 4,320 Bytes
715cc5a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
from __future__ import annotations

import argparse
from pathlib import Path
from typing import Any

from src.data.io_utils import read_jsonl


def compact(text: Any, limit: int = 240) -> str:
    value = " ".join(str(text or "").split())
    if len(value) <= limit:
        return value
    clipped = value[:limit].rsplit(" ", 1)[0].strip()
    return f"{clipped}..."


def render_prediction_case(row: dict[str, Any]) -> list[str]:
    lines = [
        f"### {row.get('category', '')}: {row.get('id', '')}",
        f"- Claim: {row.get('claim', '')}",
        f"- Gold / baseline / WikiKG / alternate: `{row.get('gold', '')}` / `{row.get('baseline_prediction', '')}` / `{row.get('wikikg_prediction', '')}` / `{row.get('alternate_prediction', '')}`",
        f"- Verified support: `{row.get('num_verified_facts', 0)}` facts, `{row.get('num_verified_triples', 0)}` triples",
    ]
    path_summary = row.get("path_summary", {})
    if path_summary:
        lines.append(
            "- Path summary: "
            f"max_final={path_summary.get('max_final_score', '')}, "
            f"max_kg_path={path_summary.get('max_kg_path_score', '')}, "
            f"max_provenance={path_summary.get('max_provenance_confidence', '')}"
        )
    if row.get("top_evidence"):
        lines.append("- Top evidence:")
        for item in row["top_evidence"][:3]:
            lines.append(f"  - [{item.get('candidate_id', '')}] {compact(item.get('text', ''))}")
    if row.get("top_verified_paths"):
        lines.append("- Top verified paths:")
        for item in row["top_verified_paths"][:3]:
            lines.append(f"  - {item.get('path_text', '')}")
            lines.append(f"    - Source: {compact(item.get('source_text', ''))}")
    if row.get("top_unsupported_triples"):
        lines.append("- Unsupported triples:")
        for item in row["top_unsupported_triples"][:2]:
            lines.append(f"  - {item.get('path_text', '')} [{item.get('nli_label', '')}]")
    lines.append("")
    return lines


def render_relation_case(row: dict[str, Any]) -> list[str]:
    lines = [
        f"### {row.get('category', '')}: {row.get('id', '')}",
        f"- Claim: {row.get('claim', '')}",
        f"- Gold / baseline / WikiKG: `{row.get('gold', '')}` / `{row.get('baseline_prediction', '')}` / `{row.get('wikikg_prediction', '')}`",
        f"- Relation: `{row.get('relation_original', '') or row.get('relation', '')}` -> `{row.get('relation', '')}`",
        f"- NLI / entailment: `{row.get('nli_label', '')}` / `{row.get('entailment_score', '')}`",
        f"- Triple: {row.get('verbalized_triple', '')}",
        f"- Source: {compact(row.get('source_text', ''), limit=320)}",
        "",
    ]
    return lines


def render_section(title: str, rows: list[dict[str, Any]], relation_mode: bool = False) -> list[str]:
    lines = [f"## {title}", ""]
    if not rows:
        lines.append("No cases selected.")
        lines.append("")
        return lines
    current_category = None
    for row in rows:
        if row.get("category") != current_category:
            current_category = row.get("category")
            lines.append(f"### Group: {current_category}")
            lines.append("")
        lines.extend(render_relation_case(row) if relation_mode else render_prediction_case(row))
    return lines


def main() -> None:
    parser = argparse.ArgumentParser()
    parser.add_argument("--averitec", type=Path, required=True)
    parser.add_argument("--healthver", type=Path, required=True)
    parser.add_argument("--vifactcheck", type=Path, required=True)
    parser.add_argument("--output", type=Path, default=Path("outputs/analysis/case_studies.md"))
    args = parser.parse_args()

    averitec_rows = read_jsonl(args.averitec)
    healthver_rows = read_jsonl(args.healthver)
    vifactcheck_rows = read_jsonl(args.vifactcheck)

    lines = ["# Stage 11 Case Studies", ""]
    lines.extend(render_section("AVeriTeC", averitec_rows))
    lines.extend(render_section("HealthVer", healthver_rows, relation_mode=True))
    lines.extend(render_section("ViFactCheck", vifactcheck_rows))
    args.output.parent.mkdir(parents=True, exist_ok=True)
    args.output.write_text("\n".join(lines).strip() + "\n", encoding="utf-8")
    print(f"Wrote case studies to {args.output}")


if __name__ == "__main__":
    main()