File size: 1,320 Bytes
56b2574 1742d27 56b2574 1742d27 56b2574 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 | {
"default": {
"description": "ACI-Bench (Attention Clarity Index Bench): a 5,000-item diagnostic benchmark measuring whether small language models (1M\u2013150M parameters) use attention \u00d7 gradient interactions to prioritize relevant context over distractors.",
"citation": "",
"homepage": "",
"license": "MIT",
"features": {
"id": {"dtype": "string", "_type": "Value"},
"topic": {"dtype": "string", "_type": "Value"},
"topicLabel": {"dtype": "string", "_type": "Value"},
"templateId": {"dtype": "string", "_type": "Value"},
"difficulty": {"dtype": "string", "_type": "Value"},
"question": {"dtype": "string", "_type": "Value"},
"referenceAnswer": {"dtype": "string", "_type": "Value"},
"context": {"dtype": "string", "_type": "Value"},
"segments": {"dtype": "string", "_type": "Value"},
"relatedSegmentIds": {"dtype": "string", "_type": "Value"},
"unrelatedSegmentIds": {"dtype": "string", "_type": "Value"},
"keyLinkPairs": {"dtype": "string", "_type": "Value"},
"meta": {"dtype": "string", "_type": "Value"}
},
"splits": {
"test": {
"name": "test",
"num_bytes": 2147483,
"num_examples": 5000,
"dataset_name": "aci-bench"
}
},
"version": "1.0.0"
}
}
|