docs
evals
scenario-1
scenario-10
scenario-2
scenario-3
scenario-4
scenario-5
scenario-6
scenario-7
scenario-8
scenario-9
{
"context": "This criteria evaluates how well the engineer uses the chardet package's SBCS n-gram analysis techniques, specifically focusing on byte normalization, n-gram extraction, binary search pattern matching, and confidence scoring mechanisms that are core to single-byte character set language detection.",
"type": "weighted_checklist",
"checklist": [
{
"name": "Byte Normalization",
"description": "Implements byte-level text normalization by converting uppercase bytes to lowercase and control characters to spaces, similar to chardet's ByteMap normalization approach used in SBCS recognizers",
"max_score": 25
},
{
"name": "N-gram Extraction",
"description": "Extracts 3-byte sequences (trigrams) from normalized text and represents them as 24-bit integer values using bit masking (0xFFFFFF), matching chardet's NGramParser implementation",
"max_score": 25
},
{
"name": "Binary Search",
"description": "Uses binary search algorithm to lookup n-grams in a sorted 64-entry pattern list, achieving O(log n) time complexity as implemented in chardet's unrolled binary search for SBCS detection",
"max_score": 30
},
{
"name": "Confidence Scoring",
"description": "Calculates confidence score based on the ratio of matched n-grams to total n-grams, applying appropriate scaling (e.g., hitCount/ngramCount × 300) and capping at 100, following chardet's SBCS confidence logic",
"max_score": 20
}
]
}