Hugging Face's logo

Spaces:
adelevett
/
ELBench-live
Sleeping

App Files Community
ELBench-live
File size: 7,621 Bytes
5db30ac
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
from __future__ import annotations

import ast
import json
import re
from difflib import SequenceMatcher
from typing import Any


def normalize_text(text: str | None) -> str:
    if text is None:
        return ""
    return re.sub(r"\s+", " ", str(text)).strip().lower()


def extract_choice_letters(text: str | None) -> list[str]:
    if not text:
        return []
    upper = str(text).upper()
    letters = re.findall(r"\b([A-Z])\b", upper)
    filtered = [letter for letter in letters if letter in {"A", "B", "C", "D", "E", "F", "G"}]
    deduped = []
    for letter in filtered:
        if letter not in deduped:
            deduped.append(letter)
    return deduped


def extract_single_choice_answer(
    text: str | None,
    valid_letters: list[str],
    *,
    answer_prefixes: list[str],
) -> str | None:
    if not text or not valid_letters or not answer_prefixes:
        return None

    lines = [line.strip() for line in str(text).splitlines() if line.strip()]
    if not lines:
        return None

    valid_pattern = "".join(re.escape(letter) for letter in valid_letters)
    candidate_lines = [_normalize_choice_line(line) for line in lines[-3:]]
    normalized_prefixes = [_normalize_choice_line(prefix) for prefix in answer_prefixes if prefix]

    for line in reversed(candidate_lines):
        if not line:
            continue
        for prefix in normalized_prefixes:
            if not prefix:
                continue
            match = re.fullmatch(
                rf"{re.escape(prefix)}\s*[:：]?\s*[\[\(οΌˆγ€γ€Œγ€Ž]?\s*([{valid_pattern}])\s*[\]\)）】」』]?\s*[.γ€‚οΌŽ!！?？]*",
                line,
                flags=re.I,
            )
            if match:
                return match.group(1).upper()
            loose_match = re.fullmatch(
                rf"{re.escape(prefix)}\s*[:：]?\s*[\[\(οΌˆγ€γ€Œγ€Ž]?\s*([{valid_pattern}])"
                rf"(?:\s*[\]\)）】」』]|[.γ€‚οΌŽγ€,，:：\-]|\s+).+",
                line,
                flags=re.I,
            )
            if loose_match:
                return loose_match.group(1).upper()

    last_line = candidate_lines[-1]
    fallback = re.fullmatch(
        rf"[\[\(οΌˆγ€γ€Œγ€Ž]?\s*([{valid_pattern}])\s*[\]\)）】」』]?\s*[.γ€‚οΌŽ!！?？]*",
        last_line,
        flags=re.I,
    )
    if fallback:
        return fallback.group(1).upper()
    return None


def extract_choice_answer(
    text: str | None,
    valid_letters: list[str],
    *,
    answer_prefixes: list[str],
    allow_multiple: bool = False,
) -> str | None:
    if not allow_multiple:
        return extract_single_choice_answer(text, valid_letters, answer_prefixes=answer_prefixes)
    if not text or not valid_letters or not answer_prefixes:
        return None

    lines = [line.strip() for line in str(text).splitlines() if line.strip()]
    if not lines:
        return None

    valid_pattern = "".join(re.escape(letter) for letter in valid_letters)
    candidate_lines = [_normalize_choice_line(line) for line in lines[-3:]]
    normalized_prefixes = [_normalize_choice_line(prefix) for prefix in answer_prefixes if prefix]

    for line in reversed(candidate_lines):
        if not line:
            continue
        for prefix in normalized_prefixes:
            if not prefix:
                continue
            match = re.fullmatch(
                rf"{re.escape(prefix)}\s*[:：]?\s*[\[\(οΌˆγ€γ€Œγ€Ž]?\s*([{valid_pattern}](?:\s*[,οΌŒγ€/]\s*[{valid_pattern}]|\s*[{valid_pattern}])*)\s*[\]\)）】」』]?\s*[.γ€‚οΌŸοΌ!?]*",
                line,
                flags=re.I,
            )
            if match:
                return _normalize_choice_answer(match.group(1), valid_letters)

    last_line = candidate_lines[-1]
    fallback = re.fullmatch(
        rf"[\[\(οΌˆγ€γ€Œγ€Ž]?\s*([{valid_pattern}](?:\s*[,οΌŒγ€/]\s*[{valid_pattern}]|\s*[{valid_pattern}])*)\s*[\]\)）】」』]?\s*[.γ€‚οΌŸοΌ!?]*",
        last_line,
        flags=re.I,
    )
    if fallback:
        return _normalize_choice_answer(fallback.group(1), valid_letters)
    return None


def normalize_choice_answer(value: str | None, valid_letters: list[str]) -> str | None:
    if value in (None, ""):
        return None
    return _normalize_choice_answer(str(value), valid_letters)


def _normalize_choice_answer(value: str, valid_letters: list[str]) -> str | None:
    valid = {letter.upper() for letter in valid_letters}
    letters = [letter for letter in re.findall(r"[A-Z]", value.upper()) if letter in valid]
    if not letters:
        return None
    deduped: list[str] = []
    for letter in letters:
        if letter not in deduped:
            deduped.append(letter)
    return "".join(deduped)


def _normalize_choice_line(text: str) -> str:
    normalized = str(text).translate(
        str.maketrans(
            {
                "：": ":",
                "（": "(",
                "οΌ‰": ")",
                "【": "[",
                "】": "]",
                "οΌ»": "[",
                "οΌ½": "]",
                "γ€Œ": "[",
                "」": "]",
                "γ€Ž": "[",
                "』": "]",
                "。": ".",
                "．": ".",
                "，": ",",
            }
        )
    )
    normalized = re.sub(r"[*_`#>\-]+", " ", normalized)
    normalized = re.sub(r"\s+", " ", normalized)
    return normalized.strip()


def parse_score_value(value: Any) -> tuple[float | None, float | None]:
    if value is None:
        return None, None
    text = str(value).strip()
    if not text:
        return None, None
    if "/" in text:
        left, right = text.split("/", 1)
        try:
            score = float(left.strip())
            total = float(right.strip())
            return score, total
        except ValueError:
            return None, None
    try:
        score = float(text)
        return score, None
    except ValueError:
        return None, None


def text_similarity(left: str | None, right: str | None) -> float:
    a = normalize_text(left)
    b = normalize_text(right)
    if not a and not b:
        return 1.0
    if not a or not b:
        return 0.0
    return SequenceMatcher(a=a, b=b).ratio()


def extract_json_object(text: str | None) -> dict[str, Any] | None:
    if not text:
        return None
    candidate = str(text).strip()
    fenced = re.findall(r"```(?:json)?\s*(\{.*?\})\s*```", candidate, flags=re.S | re.I)
    if fenced:
        candidate = fenced[0]
    else:
        match = re.search(r"\{.*\}", candidate, flags=re.S)
        if match:
            candidate = match.group(0)
    for parser in (_try_json, _try_literal_eval):
        parsed = parser(candidate)
        if isinstance(parsed, dict):
            return parsed
    return None


def extract_required_json_keys(prompt: str) -> list[str]:
    matches = re.findall(r'["\']([^"\']{1,80})["\']', prompt)
    keys: list[str] = []
    for match in matches:
        normalized = match.strip()
        if not normalized:
            continue
        if "\n" in normalized:
            continue
        if len(normalized) > 60:
            continue
        if normalized not in keys:
            keys.append(normalized)
    return keys


def _try_json(text: str) -> dict[str, Any] | None:
    try:
        parsed = json.loads(text)
        return parsed if isinstance(parsed, dict) else None
    except Exception:
        return None


def _try_literal_eval(text: str) -> dict[str, Any] | None:
    try:
        parsed = ast.literal_eval(text)
        return parsed if isinstance(parsed, dict) else None
    except Exception:
        return None