Ablation: synthetic k=1 seed=123

Ablation: synthetic k=1 seed=42

Ablation: synthetic k=0 seed=1337

Ablation: synthetic k=0 seed=789

ablation v4: mpnn_graph_train (5/5 seeds)

Ablation: synthetic k=0 seed=456

Ablation: synthetic k=0 seed=123

Ablation: synthetic k=0 seed=42

Ablation: k=20 seed=1337

Ablation: k=20 seed=789

Ablation: k=20 seed=456

Ablation: k=20 seed=123

Ablation: k=20 seed=42

ablation v4: mpnn_graph_train (4/5 seeds)

Ablation: k=10 seed=1337

Ablation: k=10 seed=789

Ablation: k=10 seed=456

Ablation: k=10 seed=123

Ablation: k=10 seed=42

Ablation: k=5 seed=1337

Ablation: k=5 seed=789

Ablation: k=5 seed=456

Ablation: k=5 seed=123

ablation v4: mpnn_graph_train (3/5 seeds)

Ablation: k=5 seed=42

Ablation: k=3 seed=1337

Ablation: k=3 seed=789

Ablation: k=3 seed=456

Ablation: k=3 seed=123

Ablation: k=3 seed=42

Ablation: k=2 seed=1337

Ablation: k=2 seed=789

Ablation: k=2 seed=456

ablation v4: mpnn_graph_train (2/5 seeds)

Ablation: k=2 seed=123

Ablation: k=2 seed=42

Ablation: k=1 seed=1337

Ablation: k=1 seed=789

Ablation: k=1 seed=456

Ablation: k=1 seed=123

MVP benchmark v2: loss curves

MVP benchmark v2: detailed per-matrix results

MVP benchmark v2: results (5 seeds)

MVP v2: suitesparse eval after seed=1337

Ablation: k=1 seed=42

MVP v2: suitesparse eval after seed=789

Ablation: k=0 seed=1337

MVP v2: suitesparse eval after seed=456

Ablation: k=0 seed=789

MVP v2: suitesparse eval after seed=123