File size: 4,520 Bytes
6835659 | 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 | from __future__ import annotations
import json
from dataclasses import dataclass
from pathlib import Path
from typing import Any, Dict, Optional
def _safe_get(d: Dict[str, Any], *keys: str, default=None):
cur: Any = d
for k in keys:
if not isinstance(cur, dict) or k not in cur:
return default
cur = cur[k]
return cur
def load_coherence_stats(path: str = "artifacts/coherence_stats.json") -> Dict[str, Any]:
p = Path(path)
if not p.exists():
return {}
try:
return json.loads(p.read_text(encoding="utf-8"))
except Exception:
return {}
def normalize_metric(stats: Dict[str, Any], name: str, value: float) -> float:
"""
Normalize metric to [0,1] using robust percentiles if available.
Expected stats formats we support:
stats[name]["p05"], stats[name]["p95"]
OR stats["metrics"][name]["p05"], ...
Falls back to (value) clipped to [0,1] if no stats found.
"""
p05 = _safe_get(stats, name, "p05")
p95 = _safe_get(stats, name, "p95")
if p05 is None or p95 is None:
p05 = _safe_get(stats, "metrics", name, "p05")
p95 = _safe_get(stats, "metrics", name, "p95")
if p05 is None or p95 is None or p95 == p05:
return max(0.0, min(1.0, float(value)))
v = (float(value) - float(p05)) / (float(p95) - float(p05))
return max(0.0, min(1.0, v))
@dataclass
class CoherenceScoringConfig:
w_msci: float = 0.35
w_st_i: float = 0.20
w_st_a: float = 0.20
w_si_a: float = 0.25
global_drift_penalty: float = 0.18
visual_drift_penalty: float = 0.10
audio_drift_penalty: float = 0.10
weakness_floor: float = 0.35
weakness_max_extra: float = 0.12
def compute_base_score(
scores: Dict[str, float],
stats: Dict[str, Any],
cfg: CoherenceScoringConfig,
) -> Dict[str, Any]:
msci = normalize_metric(stats, "msci", scores.get("msci", 0.0))
st_i = normalize_metric(stats, "st_i", scores.get("st_i", 0.0))
st_a = normalize_metric(stats, "st_a", scores.get("st_a", 0.0))
si_a = normalize_metric(stats, "si_a", scores.get("si_a", 0.0))
weights = [cfg.w_msci, cfg.w_st_i, cfg.w_st_a, cfg.w_si_a]
wsum = sum(weights) if sum(weights) > 0 else 1.0
w_msci, w_st_i, w_st_a, w_si_a = [w / wsum for w in weights]
base = w_msci * msci + w_st_i * st_i + w_st_a * st_a + w_si_a * si_a
return {
"base_score": float(max(0.0, min(1.0, base))),
"normalized": {"msci": msci, "st_i": st_i, "st_a": st_a, "si_a": si_a},
"weights": {"msci": w_msci, "st_i": w_st_i, "st_a": w_st_a, "si_a": w_si_a},
}
def compute_drift_penalties(
normalized: Dict[str, float],
drift: Dict[str, bool],
cfg: CoherenceScoringConfig,
) -> Dict[str, Any]:
penalties: Dict[str, float] = {}
if drift.get("global_drift", False):
penalties["global_drift"] = cfg.global_drift_penalty
if drift.get("visual_drift", False):
penalties["visual_drift"] = cfg.visual_drift_penalty
if drift.get("audio_drift", False):
penalties["audio_drift"] = cfg.audio_drift_penalty
weakest = min(normalized.values()) if normalized else 1.0
if weakest < cfg.weakness_floor:
ratio = (cfg.weakness_floor - weakest) / max(1e-6, cfg.weakness_floor)
penalties["weakest_modality"] = float(
min(cfg.weakness_max_extra, cfg.weakness_max_extra * ratio)
)
total = float(sum(penalties.values()))
return {"penalties": penalties, "total_penalty": total, "weakest": float(weakest)}
def compute_final_coherence(
scores: Dict[str, float],
drift: Dict[str, bool],
stats_path: str = "artifacts/coherence_stats.json",
cfg: Optional[CoherenceScoringConfig] = None,
) -> Dict[str, Any]:
cfg = cfg or CoherenceScoringConfig()
stats = load_coherence_stats(stats_path)
base_pack = compute_base_score(scores, stats, cfg)
drift_pack = compute_drift_penalties(base_pack["normalized"], drift, cfg)
final = base_pack["base_score"] - drift_pack["total_penalty"]
final = float(max(0.0, min(1.0, final)))
return {
"base_score": base_pack["base_score"],
"final_score": final,
"normalized": base_pack["normalized"],
"weights": base_pack["weights"],
"penalties": drift_pack["penalties"],
"total_penalty": drift_pack["total_penalty"],
"weakest_modality": drift_pack["weakest"],
"used_stats_file": stats_path,
"stats_loaded": bool(stats),
}
|