Datasets:
Tasks:
Text Classification
Modalities:
Text
Formats:
json
Size:
< 1K
Tags:
rag
retrieval-augmented-generation
prompt-injection
hallucination-detection
llm-evaluation
guardrails
License:
Download provenance/test_split.json from NagaYu/sieve-eval: direct link, hf CLI and curl.
- Browser
- Download file 416 Bytes
-
https://huggingface.co/datasets/NagaYu/sieve-eval/resolve/main/provenance/test_split.json
- Command line
-
hf download hf://datasets/NagaYu/sieve-eval/provenance/test_split.json
-
curl -L -o test_split.json https://huggingface.co/datasets/NagaYu/sieve-eval/resolve/main/provenance/test_split.json
416 Bytes
| { | |
| "engine_sha256": "a0d7eb43c4a40b98248bc28192226175a3978d281438d38a5107ae0f68b78b59", | |
| "dropped": [], | |
| "kept": 210, | |
| "written": { | |
| "chunk_safety": 110, | |
| "rag_diagnosis": 100 | |
| }, | |
| "note_ja_attacks": "A Japanese attack-writing batch was stopped by a safety classifier and was not retried; Japanese chunk_safety test rows are benign only (false-alarm measurement). Japanese attacks are covered by the dev split." | |
| } | |