File size: 8,241 Bytes
09f030b | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 | import argparse
import csv
import json
from pathlib import Path
import numpy as np
ROOT = Path(__file__).resolve().parents[1]
DEFAULT_MODELS = [
"qwen3.6-max",
"qwen_plus",
"deepseek_v3",
"deepseek_r1",
"llama4_scout",
]
DEFAULT_SETTINGS = ["raw_logs_context", "full_kb"]
MODEL_DISPLAY = {
"qwen3.6-max": "Qwen3.6-Max",
"qwen_plus": "Qwen-Plus",
"deepseek_v3": "DeepSeek-V3",
"deepseek_r1": "DeepSeek-R1",
"llama4_scout": "Llama4-Scout",
}
SETTING_DISPLAY = {
"raw_logs_context": "w/o KB",
"static_kb": "w/ Static KB",
"full_kb": "w/ KB",
}
def load_json(path):
path = Path(path)
if not path.exists():
raise FileNotFoundError(path)
return json.loads(path.read_text(encoding="utf-8"))
def mean(values):
return float(np.mean(values)) if values else 0.0
def summarize_file(path):
data = load_json(path)
required_scores = {
"correctness": [],
"contextual_relevance": [],
"clarity": [],
}
all_scores = []
case_counts = {"TP": 0, "FP": 0, "TN": 0, "FN": 0, "ERROR": 0}
for item in data:
evaluation = item.get("evaluation")
if not evaluation:
continue
case_type = evaluation.get("case_type", "ERROR")
case_counts[case_type] = case_counts.get(case_type, 0) + 1
all_scores.append(float(evaluation.get("average_score", 0)))
# Required response quality is computed over the fixed set of
# ground-truth proactive-required cases. TP uses the judge score; FN is
# already scored as 0 by quality_assessment.py. This avoids TP-only
# selection bias when different settings recall different instances.
if item.get("gt_proactive") is True:
required_scores["correctness"].append(float(evaluation.get("correctness", 0)))
required_scores["contextual_relevance"].append(
float(evaluation.get("contextual_relevance", evaluation.get("contextual_alignment", 0)))
)
required_scores["clarity"].append(float(evaluation.get("clarity", 0)))
total = sum(case_counts.values())
return {
"records": total,
"required_cases": len(required_scores["correctness"]),
"tp": case_counts.get("TP", 0),
"fp": case_counts.get("FP", 0),
"tn": case_counts.get("TN", 0),
"fn": case_counts.get("FN", 0),
"errors": case_counts.get("ERROR", 0),
"correctness": mean(required_scores["correctness"]),
"contextual_relevance": mean(required_scores["contextual_relevance"]),
"clarity": mean(required_scores["clarity"]),
"overall_effectiveness": mean(all_scores),
}
def write_csv(rows, output_path):
output_path.parent.mkdir(parents=True, exist_ok=True)
fieldnames = [
"model",
"model_display",
"setting",
"setting_display",
"records",
"required_cases",
"TP",
"FP",
"TN",
"FN",
"errors",
"correctness",
"contextual_relevance",
"clarity",
"overall_effectiveness",
]
with output_path.open("w", encoding="utf-8", newline="") as f:
writer = csv.DictWriter(f, fieldnames=fieldnames)
writer.writeheader()
writer.writerows(rows)
def latex_rows(rows):
by_model = {}
for row in rows:
by_model.setdefault(row["model"], {})[row["setting"]] = row
def fmt_pair(raw_value, full_value):
raw_text = f"{raw_value:.2f}"
full_text = f"{full_value:.2f}"
if full_value > raw_value:
full_text = f"\\textbf{{{full_text}}}"
elif raw_value > full_value:
raw_text = f"\\textbf{{{raw_text}}}"
return raw_text, full_text
lines = []
for model in DEFAULT_MODELS:
model_rows = by_model.get(model, {})
raw = model_rows.get("raw_logs_context")
full = model_rows.get("full_kb")
if not raw or not full:
continue
gain = full["overall_effectiveness"] - raw["overall_effectiveness"]
display = MODEL_DISPLAY.get(model, model)
raw_correctness, full_correctness = fmt_pair(raw["correctness"], full["correctness"])
raw_contextual, full_contextual = fmt_pair(
raw["contextual_relevance"], full["contextual_relevance"]
)
raw_clarity, full_clarity = fmt_pair(raw["clarity"], full["clarity"])
raw_overall, full_overall = fmt_pair(
raw["overall_effectiveness"], full["overall_effectiveness"]
)
lines.append(f"% --- {display} ---")
lines.append(f"\\multirow{{2}}{{*}}{{{display}}}")
lines.append(
" & w/o KB "
f"& {raw_correctness} & {raw_contextual} "
f"& {raw_clarity} & {raw_overall} \\\\"
)
lines.append(
" & w/ KB "
f"& {full_correctness} "
f"& {full_contextual} "
f"& {full_clarity} "
f"& {full_overall}\\inc{{{gain:.2f}}} \\\\"
)
lines.append("\\cmidrule(lr){1-6}")
lines.append("")
return "\n".join(lines)
def parse_args():
parser = argparse.ArgumentParser(description="Summarize proactive quality scores.")
parser.add_argument("--output-root", type=Path, default=ROOT / "results")
parser.add_argument(
"--score-dir",
default="quality_score_strict",
help="Directory under output-root containing scored JSON files.",
)
parser.add_argument("--models", nargs="+", default=DEFAULT_MODELS)
parser.add_argument("--settings", nargs="+", default=DEFAULT_SETTINGS)
parser.add_argument(
"--csv",
type=Path,
default=None,
help="Optional output CSV path. Defaults to <output-root>/metrics/quality_effectiveness.csv.",
)
parser.add_argument(
"--latex",
type=Path,
default=None,
help="Optional LaTeX row output. Defaults to <output-root>/metrics/quality_effectiveness_latex.md.",
)
return parser.parse_args()
def main():
args = parse_args()
rows = []
for setting in args.settings:
for model in args.models:
path = args.output_root / args.score_dir / setting / f"experiment_{model}_{setting}_scored.json"
if not path.exists():
print(f"Missing scored file: {path}")
continue
summary = summarize_file(path)
rows.append(
{
"model": model,
"model_display": MODEL_DISPLAY.get(model, model),
"setting": setting,
"setting_display": SETTING_DISPLAY.get(setting, setting),
"records": summary["records"],
"required_cases": summary["required_cases"],
"TP": summary["tp"],
"FP": summary["fp"],
"TN": summary["tn"],
"FN": summary["fn"],
"errors": summary["errors"],
"correctness": round(summary["correctness"], 2),
"contextual_relevance": round(summary["contextual_relevance"], 2),
"clarity": round(summary["clarity"], 2),
"overall_effectiveness": round(summary["overall_effectiveness"], 2),
}
)
csv_path = args.csv or args.output_root / "metrics" / "quality_effectiveness.csv"
latex_path = args.latex or args.output_root / "metrics" / "quality_effectiveness_latex.md"
write_csv(rows, csv_path)
latex_path.parent.mkdir(parents=True, exist_ok=True)
latex_text = latex_rows(rows)
latex_path.write_text(latex_text + "\n", encoding="utf-8")
print(f"Wrote CSV: {csv_path}")
print(f"Wrote LaTeX rows: {latex_path}")
print()
for row in rows:
print(
f"{row['model_display']} / {row['setting_display']}: "
f"C={row['correctness']:.2f}, Contextual Relevance={row['contextual_relevance']:.2f}, "
f"Clarity={row['clarity']:.2f}, Overall={row['overall_effectiveness']:.2f}, "
f"errors={row['errors']}"
)
if __name__ == "__main__":
main()
|