Download context_scaling/protocol.json from wangyanbo/MindView-Bench: direct link, hf CLI and curl.
- Browser
- Download file 742 Bytes
-
https://huggingface.co/datasets/wangyanbo/MindView-Bench/resolve/main/context_scaling/protocol.json
- Command line
-
hf download hf://datasets/wangyanbo/MindView-Bench/context_scaling/protocol.json
-
curl -L -o protocol.json https://huggingface.co/datasets/wangyanbo/MindView-Bench/resolve/main/context_scaling/protocol.json
742 Bytes
| { | |
| "schema_version": "mindview_context_scaling_v1", | |
| "index": "context_scaling/metadata.jsonl", | |
| "index_sha256": "97be55cb44c45d693acd370175805804c12f95b99bebb115f0df1e1b2077219b", | |
| "records": 4500, | |
| "paired_cases": 1500, | |
| "configurations": { | |
| "c3": 3, | |
| "c6": 6, | |
| "c10": 10 | |
| }, | |
| "pair_index": "context_scaling/pairs.jsonl", | |
| "pairing": "Use the explicit mapping: original sample IDs differ across configurations.", | |
| "invariants": [ | |
| "Same query, task, scene, target, and anchor across each triplet", | |
| "Ordered nested input-view sets", | |
| "Target excluded from all input-view sets" | |
| ], | |
| "scoring": "Report each configuration separately; fixed ten-view geometry scoring references are shared across paired cases." | |
| } | |