File size: 4,581 Bytes
b3b12cd | 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 | """Test B+S sub-composite against every available external proxy."""
import csv, json, math, sqlite3
from pathlib import Path
import sys
sys.path.insert(0, "/tmp/AgentPulse-Data-v6/code")
from reproduce_table3 import (load_latest_scores, load_latest_signals, load_registry,
benchmark_sentiment_subcomposite, spearman, two_sided_p)
csv_dir = Path("/tmp/AgentPulse-Data-v6/data/csv")
scores = load_latest_scores(csv_dir)
signals = load_latest_signals(csv_dir)
registry = load_registry(csv_dir)
def factors(name):
s = scores[name]
sub = json.loads(s.get("sub_scores") or "[0.5,0,0]")
b = sub[0] if sub else 0.5
sent = float(s.get("sentiment") or 0)
sn = max(0, min(1, (sent+1)/2))
return b, sn
def bs(name):
b, s = factors(name)
return 0.6364*b + 0.3636*s
def b_only(name):
return factors(name)[0]
# Pull pypi history aggregates
conn = sqlite3.connect("/tmp/AgentPulse-Data-v6/data/sqlite/arena_agent_subset.db")
pypi_total = {}; pypi_recent_30 = {}
for r in conn.execute("SELECT agent_name, SUM(downloads) FROM agent_pypi_history GROUP BY agent_name"):
pypi_total[r[0]] = r[1]
for r in conn.execute("SELECT agent_name, SUM(downloads) FROM agent_pypi_history WHERE date >= '2026-04-01' GROUP BY agent_name"):
pypi_recent_30[r[0]] = r[1]
conn.close()
# Eligibility: agents with B+S computable + github_repo
eligible_all = [n for n in registry if registry[n].get("github_repo") and n in scores and n in signals]
print(f"Total eligible: {len(eligible_all)}\n")
PROXIES = [
# (label, value_fn, log_transform, requires_nonzero)
("vscode_installs", lambda n: signals[n].get("vscode_installs", 0), True, True),
("vscode_installs (zero-impute)", lambda n: signals[n].get("vscode_installs", 0), True, False),
("vscode_rating", lambda n: signals[n].get("vscode_rating", 0), False, True),
("vscode_rating_count", lambda n: signals[n].get("vscode_rating_count", 0), True, True),
("github_stars", lambda n: signals[n].get("github_stars", 0), True, True),
("github_forks", lambda n: signals[n].get("github_forks", 0), True, True),
("github_open_issues", lambda n: signals[n].get("github_open_issues", 0), True, True),
("github_watchers", lambda n: signals[n].get("github_watchers", 0), True, True),
("github_contributors", lambda n: signals[n].get("github_contributors", 0), True, True),
("issue_close_rate", lambda n: signals[n].get("issue_close_rate", 0), False, True),
("days_since_update (inv)", lambda n: -1 * signals[n].get("days_since_update", 0), False, True),
("pypi_downloads_week", lambda n: signals[n].get("pypi_downloads_week", 0), True, True),
("pypi_downloads_month", lambda n: signals[n].get("pypi_downloads_month", 0), True, True),
("pypi_history_total", lambda n: pypi_total.get(n, 0), True, True),
("pypi_history_recent30", lambda n: pypi_recent_30.get(n, 0), True, True),
("so_questions", lambda n: signals[n].get("so_questions", 0), False, False),
("so_questions (log)", lambda n: signals[n].get("so_questions", 0), True, False),
("mention_count", lambda n: int(scores[n].get("mention_count") or 0), True, False),
]
def correlate_with(sub_name, sub_fn):
print(f"\n{'='*100}")
print(f" Sub-composite: {sub_name}")
print(f"{'='*100}")
print(f" {'proxy':<40} {'rho_s':>7} {'p':>8} {'n':>4} notes")
print(f" {'-'*40} {'-'*7} {'-'*8} {'-'*4}")
for label, fn, log_xform, require_nonzero in PROXIES:
elig = []
for n in eligible_all:
v = fn(n)
if require_nonzero and (v is None or v == 0): continue
if v is None: v = 0
elig.append((n, v))
if len(elig) < 4:
print(f" {label:<40} (n={len(elig)} too small)")
continue
xs = [sub_fn(n) for n,_ in elig]
ys = [(math.log10(v+1) if log_xform else v) for _,v in elig]
rho = spearman(xs, ys); p = two_sided_p(rho, len(elig))
sig = "***" if p<0.001 else "** " if p<0.01 else "* " if p<0.05 else " "
note = ""
if rho > 0.3 and p < 0.05: note = "<-- candidate"
if abs(rho) < 0.10: note = "(null)"
print(f" {label:<40} {rho:+7.3f} {p:>8.4f} {len(elig):>4} {sig}{note}")
correlate_with("B+S (paper's pre-specified)", bs)
correlate_with("B-only (capability alone)", b_only)
|