Add real sha256; remove non-data FileObjects; add rai:personalSensitiveInformation + rai:hasSyntheticData
Browse files- croissant.json +33 -25
croissant.json
CHANGED
|
@@ -60,29 +60,14 @@
|
|
| 60 |
"inLanguage": "en",
|
| 61 |
"citeAs": "@inproceedings{anonymous2026tabmibench, title={TabMI-Bench: Evaluating Mechanistic Interpretability Methods Across Tabular Foundation Model Architectures}, author={Anonymous}, booktitle={Advances in Neural Information Processing Systems (NeurIPS) Evaluations \\& Datasets Track}, year={2026}}",
|
| 62 |
"distribution": [
|
| 63 |
-
{
|
| 64 |
-
"@type": "cr:FileObject",
|
| 65 |
-
"@id": "github-repo",
|
| 66 |
-
"name": "github-repo",
|
| 67 |
-
"description": "Public GitHub repository containing all source code, hooks, evaluation scripts, tests, and frozen artifacts.",
|
| 68 |
-
"contentUrl": "https://github.com/evaldataset/TabMI-Bench",
|
| 69 |
-
"encodingFormat": "application/zip"
|
| 70 |
-
},
|
| 71 |
-
{
|
| 72 |
-
"@type": "cr:FileObject",
|
| 73 |
-
"@id": "huggingface-dataset",
|
| 74 |
-
"name": "huggingface-dataset",
|
| 75 |
-
"description": "HuggingFace dataset repository with frozen JSON aggregated artifacts driving paper-facing tables and figures.",
|
| 76 |
-
"contentUrl": "https://huggingface.co/datasets/EvalData/TabMI-Bench",
|
| 77 |
-
"encodingFormat": "application/json"
|
| 78 |
-
},
|
| 79 |
{
|
| 80 |
"@type": "cr:FileObject",
|
| 81 |
"@id": "rd5-fullscale-aggregated",
|
| 82 |
"name": "rd5-fullscale-aggregated",
|
| 83 |
"description": "Aggregated 5-seed multi-model intermediary probing results (Table 1 source).",
|
| 84 |
"contentUrl": "https://huggingface.co/datasets/EvalData/TabMI-Bench/resolve/main/rd5_fullscale_aggregated.json",
|
| 85 |
-
"encodingFormat": "application/json"
|
|
|
|
| 86 |
},
|
| 87 |
{
|
| 88 |
"@type": "cr:FileObject",
|
|
@@ -90,7 +75,8 @@
|
|
| 90 |
"name": "tabdpt-probing-3seed",
|
| 91 |
"description": "TabDPT in-family holdout probing results aggregated over 3 seeds.",
|
| 92 |
"contentUrl": "https://huggingface.co/datasets/EvalData/TabMI-Bench/resolve/main/tabdpt_probing_3seed.json",
|
| 93 |
-
"encodingFormat": "application/json"
|
|
|
|
| 94 |
},
|
| 95 |
{
|
| 96 |
"@type": "cr:FileObject",
|
|
@@ -98,7 +84,8 @@
|
|
| 98 |
"name": "tabdpt-causal-3seed",
|
| 99 |
"description": "TabDPT noising-based causal tracing results aggregated over 3 seeds.",
|
| 100 |
"contentUrl": "https://huggingface.co/datasets/EvalData/TabMI-Bench/resolve/main/tabdpt_causal_3seed.json",
|
| 101 |
-
"encodingFormat": "application/json"
|
|
|
|
| 102 |
},
|
| 103 |
{
|
| 104 |
"@type": "cr:FileObject",
|
|
@@ -106,7 +93,8 @@
|
|
| 106 |
"name": "nam-holdout",
|
| 107 |
"description": "NAM out-of-family holdout per-seed layer-wise R^2 profiles.",
|
| 108 |
"contentUrl": "https://huggingface.co/datasets/EvalData/TabMI-Bench/resolve/main/nam_holdout.json",
|
| 109 |
-
"encodingFormat": "application/json"
|
|
|
|
| 110 |
},
|
| 111 |
{
|
| 112 |
"@type": "cr:FileObject",
|
|
@@ -114,7 +102,8 @@
|
|
| 114 |
"name": "lofo-primary-endpoint",
|
| 115 |
"description": "Leave-one-function-out robustness for the primary endpoint sigma2_profile.",
|
| 116 |
"contentUrl": "https://huggingface.co/datasets/EvalData/TabMI-Bench/resolve/main/lofo_primary_endpoint.json",
|
| 117 |
-
"encodingFormat": "application/json"
|
|
|
|
| 118 |
},
|
| 119 |
{
|
| 120 |
"@type": "cr:FileObject",
|
|
@@ -122,7 +111,26 @@
|
|
| 122 |
"name": "tabpfn25-fullscale-aggregated",
|
| 123 |
"description": "TabPFN v2 vs v2.5 logit-lens, coefficient probing, intermediary probing, and CKA aggregated over 3 seeds.",
|
| 124 |
"contentUrl": "https://huggingface.co/datasets/EvalData/TabMI-Bench/resolve/main/tabpfn25_fullscale_aggregated.json",
|
| 125 |
-
"encodingFormat": "application/json"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 126 |
}
|
| 127 |
],
|
| 128 |
"recordSet": [
|
|
@@ -254,7 +262,7 @@
|
|
| 254 |
"description": "MI technique name (e.g. linear-probing, logit-lens, activation-patching, sae, cka).",
|
| 255 |
"dataType": "sc:Text",
|
| 256 |
"source": {
|
| 257 |
-
"fileObject": {"@id": "
|
| 258 |
"extract": {"jsonPath": "$.technique"}
|
| 259 |
}
|
| 260 |
},
|
|
@@ -265,7 +273,7 @@
|
|
| 265 |
"description": "TFM model name.",
|
| 266 |
"dataType": "sc:Text",
|
| 267 |
"source": {
|
| 268 |
-
"fileObject": {"@id": "
|
| 269 |
"extract": {"jsonPath": "$.model"}
|
| 270 |
}
|
| 271 |
},
|
|
@@ -276,7 +284,7 @@
|
|
| 276 |
"description": "Applicability label: Supported, Limited, or Not established.",
|
| 277 |
"dataType": "sc:Text",
|
| 278 |
"source": {
|
| 279 |
-
"fileObject": {"@id": "
|
| 280 |
"extract": {"jsonPath": "$.label"}
|
| 281 |
}
|
| 282 |
}
|
|
|
|
| 60 |
"inLanguage": "en",
|
| 61 |
"citeAs": "@inproceedings{anonymous2026tabmibench, title={TabMI-Bench: Evaluating Mechanistic Interpretability Methods Across Tabular Foundation Model Architectures}, author={Anonymous}, booktitle={Advances in Neural Information Processing Systems (NeurIPS) Evaluations \\& Datasets Track}, year={2026}}",
|
| 62 |
"distribution": [
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 63 |
{
|
| 64 |
"@type": "cr:FileObject",
|
| 65 |
"@id": "rd5-fullscale-aggregated",
|
| 66 |
"name": "rd5-fullscale-aggregated",
|
| 67 |
"description": "Aggregated 5-seed multi-model intermediary probing results (Table 1 source).",
|
| 68 |
"contentUrl": "https://huggingface.co/datasets/EvalData/TabMI-Bench/resolve/main/rd5_fullscale_aggregated.json",
|
| 69 |
+
"encodingFormat": "application/json",
|
| 70 |
+
"sha256": "8bae47fa6f777ff1b8e4d8950470d99f3a024478d43997b09f4b24642881a1c3"
|
| 71 |
},
|
| 72 |
{
|
| 73 |
"@type": "cr:FileObject",
|
|
|
|
| 75 |
"name": "tabdpt-probing-3seed",
|
| 76 |
"description": "TabDPT in-family holdout probing results aggregated over 3 seeds.",
|
| 77 |
"contentUrl": "https://huggingface.co/datasets/EvalData/TabMI-Bench/resolve/main/tabdpt_probing_3seed.json",
|
| 78 |
+
"encodingFormat": "application/json",
|
| 79 |
+
"sha256": "09e6fe1cfcdec668ee6bbc0f8927aef63af4f10039c7db28145b4646ee53f055"
|
| 80 |
},
|
| 81 |
{
|
| 82 |
"@type": "cr:FileObject",
|
|
|
|
| 84 |
"name": "tabdpt-causal-3seed",
|
| 85 |
"description": "TabDPT noising-based causal tracing results aggregated over 3 seeds.",
|
| 86 |
"contentUrl": "https://huggingface.co/datasets/EvalData/TabMI-Bench/resolve/main/tabdpt_causal_3seed.json",
|
| 87 |
+
"encodingFormat": "application/json",
|
| 88 |
+
"sha256": "a19828d2f5ca3c3812651d5179532ff3fa882a01d92cf74735cab195859cf54b"
|
| 89 |
},
|
| 90 |
{
|
| 91 |
"@type": "cr:FileObject",
|
|
|
|
| 93 |
"name": "nam-holdout",
|
| 94 |
"description": "NAM out-of-family holdout per-seed layer-wise R^2 profiles.",
|
| 95 |
"contentUrl": "https://huggingface.co/datasets/EvalData/TabMI-Bench/resolve/main/nam_holdout.json",
|
| 96 |
+
"encodingFormat": "application/json",
|
| 97 |
+
"sha256": "ebf795a470ea11e5fa5b6d2a85c6287add70494cecba81e94ba038ed2c3653af"
|
| 98 |
},
|
| 99 |
{
|
| 100 |
"@type": "cr:FileObject",
|
|
|
|
| 102 |
"name": "lofo-primary-endpoint",
|
| 103 |
"description": "Leave-one-function-out robustness for the primary endpoint sigma2_profile.",
|
| 104 |
"contentUrl": "https://huggingface.co/datasets/EvalData/TabMI-Bench/resolve/main/lofo_primary_endpoint.json",
|
| 105 |
+
"encodingFormat": "application/json",
|
| 106 |
+
"sha256": "66fe13294502cd9c0a32616861b0ccb4aa11e7544da54f60b98943ad55173419"
|
| 107 |
},
|
| 108 |
{
|
| 109 |
"@type": "cr:FileObject",
|
|
|
|
| 111 |
"name": "tabpfn25-fullscale-aggregated",
|
| 112 |
"description": "TabPFN v2 vs v2.5 logit-lens, coefficient probing, intermediary probing, and CKA aggregated over 3 seeds.",
|
| 113 |
"contentUrl": "https://huggingface.co/datasets/EvalData/TabMI-Bench/resolve/main/tabpfn25_fullscale_aggregated.json",
|
| 114 |
+
"encodingFormat": "application/json",
|
| 115 |
+
"sha256": "3f81492cd2183a5e8bc82216162c75579bffccf5c19a0cc07b7533a85db94773"
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"@type": "cr:FileObject",
|
| 119 |
+
"@id": "c1c2-baselines",
|
| 120 |
+
"name": "c1c2-baselines",
|
| 121 |
+
"description": "Negative-control baselines: shuffled-label probing and random-target probing per model and layer.",
|
| 122 |
+
"contentUrl": "https://huggingface.co/datasets/EvalData/TabMI-Bench/resolve/main/c1c2_baselines.json",
|
| 123 |
+
"encodingFormat": "application/json",
|
| 124 |
+
"sha256": "c19cd33ef8e1e7b03ad26956cff8f932d985ce4f1d234a030d619bd7a7aadf5f"
|
| 125 |
+
},
|
| 126 |
+
{
|
| 127 |
+
"@type": "cr:FileObject",
|
| 128 |
+
"@id": "scale-10k-multiseed",
|
| 129 |
+
"name": "scale-10k-multiseed",
|
| 130 |
+
"description": "N=10K probing scale validation results aggregated over 3 seeds.",
|
| 131 |
+
"contentUrl": "https://huggingface.co/datasets/EvalData/TabMI-Bench/resolve/main/scale_10k_multiseed.json",
|
| 132 |
+
"encodingFormat": "application/json",
|
| 133 |
+
"sha256": "ecf8cb8572dcd9f4ea1c6cd569b52ab9b38a2d406ab18bb1b2da766f8e853eca"
|
| 134 |
}
|
| 135 |
],
|
| 136 |
"recordSet": [
|
|
|
|
| 262 |
"description": "MI technique name (e.g. linear-probing, logit-lens, activation-patching, sae, cka).",
|
| 263 |
"dataType": "sc:Text",
|
| 264 |
"source": {
|
| 265 |
+
"fileObject": {"@id": "rd5-fullscale-aggregated"},
|
| 266 |
"extract": {"jsonPath": "$.technique"}
|
| 267 |
}
|
| 268 |
},
|
|
|
|
| 273 |
"description": "TFM model name.",
|
| 274 |
"dataType": "sc:Text",
|
| 275 |
"source": {
|
| 276 |
+
"fileObject": {"@id": "rd5-fullscale-aggregated"},
|
| 277 |
"extract": {"jsonPath": "$.model"}
|
| 278 |
}
|
| 279 |
},
|
|
|
|
| 284 |
"description": "Applicability label: Supported, Limited, or Not established.",
|
| 285 |
"dataType": "sc:Text",
|
| 286 |
"source": {
|
| 287 |
+
"fileObject": {"@id": "rd5-fullscale-aggregated"},
|
| 288 |
"extract": {"jsonPath": "$.label"}
|
| 289 |
}
|
| 290 |
}
|