MorphMind-AI/CFM-Methods-7B

Update benchmark figure (no truncation)

Refresh card: clean two-panel benchmark figure, worked example, multi-domain framing

Update weights: v2 conformance-optimized checkpoint (held-out R 0.984, FPR 0.000)

Upload benchmark.png with huggingface_hub

Add SOTA comparison: matches frontier recall/localization, 0 false alarms vs Opus 0.28 / GPT-4o 0.47

Positive MorphMind framing

CFM-Methods-7B: methodology screener (0.95 held-out recall, 0.95 loc, 0.00 FPR)

initial commit