leaderboard_diff_chart.py

#!/usr/bin/env python3
"""
Minneapolis Summit Slide — Leaderboard Diff Chart

Generates a dual-axis chart showing:
- Composite Score (ProofScore) by model (bar chart)
- Spectral Event Rate by model (line overlay)

Sorted by ProofScore descending.
"""

import json
from pathlib import Path

# Load leaderboard
leaderboard_path = Path(__file__).parent / "leaderboard.json"
with open(leaderboard_path) as f:
    data = json.load(f)

# Extract model data
models = []
for entry in data["ranking"]:
    models.append({
        "model_id": entry["model_id"],
        "composite_score": entry["composite_score"],
        "spectral_event_rate": entry["spectral_event_rate"],
        "mean_latency_ms": entry["mean_latency_ms"],
        "mean_tokens_per_second": entry["mean_tokens_per_second"]
    })

# Sort by composite_score descending
models.sort(key=lambda x: x["composite_score"], reverse=True)

# Print table for slide
print("=" * 100)
print("MINNEAPOLIS SUMMIT - MAN IN THE ARENA LEADERBOARD")
print("Full Sweep 8-Model Edition | Chain Verified: 16/16 Intact")
print("=" * 100)
print()
print(f"{'Rank':<6}{'Model':<35}{'ProofScore':<12}{'Spectral Rate':<15}{'Latency (ms)':<14}{'Tok/s':<10}")
print("-" * 100)

for i, m in enumerate(models, 1):
    print(f"{i:<6}{m['model_id']:<35}{m['composite_score']:<12.6f}{m['spectral_event_rate']:<15.4f}{m['mean_latency_ms']:<14.3f}{m['mean_tokens_per_second']:<10.3f}")

print()
print(f"Top Model: {data['chain_verification']['status']} | Merkle Root: {data['chain_verification']['merkle_root'][:32]}...")
print(f"Total Entries: {data['case_result_count']} | Models: {data['model_count']}")

# Generate ASCII chart
print()
print("=" * 100)
print("COMPOSITE SCORE COMPARISON (ProofScore)")
print("=" * 100)

max_score = max(m["composite_score"] for m in models)
bar_width = 50

for m in models:
    bar_len = int((m["composite_score"] / max_score) * bar_width) if max_score > 0 else 0
    bar = "#" * bar_len
    spectral_marker = "[STABLE]" if m["spectral_event_rate"] < 0.75 else "[MARGIN]" if m["spectral_event_rate"] < 1.0 else "[UNSTABLE]"
    print(f"{m['model_id'][:30]:<30} |{bar:<50}| {m['composite_score']:.4f} {spectral_marker}")

print()
print("=" * 100)
print("SPECTRAL EVENT RATE (Lower = More Stable)")
print("=" * 100)

for m in models:
    stability = "STABLE" if m["spectral_event_rate"] < 0.75 else "MARGINAL" if m["spectral_event_rate"] < 1.0 else "UNSTABLE"
    print(f"{m['model_id'][:35]:<35} | Rate: {m['spectral_event_rate']:.4f} | {stability}")

# Key insight
print()
print("=" * 100)
print("KEY INSIGHT")
print("=" * 100)
top_model = models[0]
bottom_models = [m for m in models if m["composite_score"] == 0.0]
print(f"Top performer: {top_model['model_id']} (8B params) with ProofScore {top_model['composite_score']:.4f}")
if bottom_models:
    print(f"Zero scorers: {len(bottom_models)} models including 26B+ parameter models")
    print()
    print("-> Koopman spectral gate filters for DYNAMICAL STABILITY, not parameter count")
    print("-> Smaller, well-aligned models outperform larger, unaligned ones")
    print("-> ProofScore = claim_verification_f1 * (1 - spectral_penalty)")