eval-results-mib-subgraph/baselines/EAP_OA.json

{
    "method_name": "EAP (OA)",
    "results": [
        {
            "model_id": "gpt2",
            "scores": {
                "ioi": {
                    "edge_counts": [
                        3.0,
                        8.0,
                        18.0,
                        66.0,
                        119.0,
                        616.0,
                        2699.0,
                        6055.0,
                        14988.0,
                        32491.0
                    ],
                    "faithfulness": [
                        0.0,
                        0.0,
                        0.0,
                        -0.02605376953427554,
                        -0.027599262798968967,
                        0.018682773460249002,
                        0.3035174289407458,
                        0.905198792680162,
                        1.2374291327745524,
                        1.0
                    ]
                }
            }
        },
        {
            "model_id": "qwen2.5",
            "scores": {
                "mcqa": {
                    "edge_counts": [
                        19.0,
                        81.0,
                        203.0,
                        360.0,
                        872.0,
                        7146.0,
                        17450.0,
                        35719.0,
                        87921.0,
                        179749.0
                    ],
                    "faithfulness": [
                        0.0,
                        -0.0031645569620253164,
                        -0.006329113924050633,
                        -0.015822784810126583,
                        0.03164556962025317,
                        0.09651898734177215,
                        0.189873417721519,
                        0.13132911392405064,
                        0.07278481012658228,
                        1.0
                    ]
                },
                "ioi": {
                    "edge_counts": [
                        7.0,
                        62.0,
                        163.0,
                        393.0,
                        868.0,
                        7103.0,
                        17098.0,
                        35450.0,
                        88180.0,
                        179749.0
                    ],
                    "faithfulness": [
                        0.0,
                        0.0,
                        0.0,
                        0.16043613707165108,
                        0.22429906542056074,
                        0.38317757009345793,
                        0.5595794392523364,
                        0.6433021806853583,
                        0.6448598130841121,
                        1.0
                    ]
                }
            }
        }
    ]
}