OpenCompass/opencompass/datasets/subjective/arena_hard.py
2025-04-11 16:59:40 +08:00

279 lines
8.7 KiB
Python

# flake8: noqa: W605
import json
import math
import os.path as osp
import re
from collections import defaultdict
import numpy as np
import pandas as pd
from datasets import Dataset, DatasetDict
from sklearn.linear_model import LogisticRegression
from opencompass.datasets.subjective.compass_arena_subjective_bench import \
get_element_counts
from opencompass.registry import DICT_POSTPROCESSORS, LOAD_DATASET
from opencompass.utils import get_data_path
from ..base import BaseDataset
from .utils import get_judgeanswer_and_reference
@LOAD_DATASET.register_module()
class ArenaHardDataset(BaseDataset):
def load(self, path: str, name: str, *args, **kwargs):
path = get_data_path(path, local_mode=True)
filename = osp.join(path, f'{name}.jsonl')
dataset = DatasetDict()
raw_data = []
with open(filename, 'r', encoding='utf-8') as file:
for line in file:
problem = json.loads(line)
question_id = problem['question_id']
cluster = problem['cluster']
question = problem['turns'][0][
'content'] # only one turn in arena_hard
raw_data.append({
'question': question,
'capability': cluster,
'judge': {
'capability': cluster,
'question': question,
'question_id': question_id,
},
})
dataset = Dataset.from_list(raw_data)
return dataset
def post_process_arenahard(completion):
s = completion['prediction']
if result := re.findall('\[\[([AB<>=]+)\]\]', s):
return result[0]
else:
return None
def get_battles_from_judgment(judged_answers, references, WEIGHT=3):
arena_hard_battles = pd.DataFrame()
for judged_answer, reference in zip(judged_answers, references):
output = {
'model_a': reference['answer1'],
'model_b': reference['answer2']
}
if judged_answer is not None:
weight = 1
if judged_answer == 'A=B':
output['winner'] = 'tie'
elif judged_answer == 'A>B':
output['winner'] = 'model_a'
elif judged_answer == 'A>>B':
output['winner'] = 'model_a'
weight = WEIGHT
elif judged_answer == 'B>A':
output['winner'] = 'model_b'
elif judged_answer == 'B>>A':
output['winner'] = 'model_b'
weight = WEIGHT
else:
weight = 0
else:
weight = 0
if weight:
arena_hard_battles = pd.concat(
[arena_hard_battles,
pd.DataFrame([output] * weight)])
return arena_hard_battles
def compute_mle_elo(df, SCALE=400, BASE=10, INIT_RATING=1000):
models = pd.concat([df['model_a'], df['model_b']]).unique()
models = pd.Series(np.arange(len(models)), index=models)
# duplicate battles
df = pd.concat([df, df], ignore_index=True)
p = len(models.index)
n = df.shape[0]
X = np.zeros([n, p])
X[np.arange(n), models[df['model_a']]] = +math.log(BASE)
X[np.arange(n), models[df['model_b']]] = -math.log(BASE)
# one A win => two A win
Y = np.zeros(n)
Y[df['winner'] == 'model_a'] = 1.0
# one tie => one A win + one B win
# find tie + tie (both bad) index
tie_idx = (df['winner'] == 'tie') | (df['winner'] == 'tie (bothbad)')
tie_idx[len(tie_idx) // 2:] = False
Y[tie_idx] = 1.0
lr = LogisticRegression(
fit_intercept=False, penalty=None, tol=1e-8
) # May need to set a small value when not use GPT4 as judge model
lr.fit(X, Y)
elo_scores = SCALE * lr.coef_[0] + INIT_RATING
# set anchor as gpt4-0314 = 1000
if 'gpt4-0314' in models.index:
elo_scores += 1000 - elo_scores[models['gpt4-0314']]
return pd.Series(elo_scores,
index=models.index).sort_values(ascending=False)
def get_bootstrap_result(battles, func_compute_elo, num_round):
rows = []
for i in range(num_round):
rows.append(func_compute_elo(battles.sample(frac=1.0, replace=True)))
df = pd.DataFrame(rows)
return df[df.median().sort_values(ascending=False).index]
def preety_print_two_ratings(ratings_1, ratings_2, column_names):
df = (pd.DataFrame(
[[n, ratings_1[n], ratings_2[n]] for n in ratings_1.keys()],
columns=['Model', column_names[0], column_names[1]],
).sort_values(column_names[0], ascending=False).reset_index(drop=True))
df[column_names[0]] = (df[column_names[0]] + 0.5).astype(int)
df[column_names[1]] = (df[column_names[1]] + 0.5).astype(int)
df.index = df.index + 1
return df
def predict_win_rate(elo_ratings, SCALE=400, BASE=10, INIT_RATING=1000):
names = list(elo_ratings.keys())
wins = defaultdict(lambda: defaultdict(lambda: 0))
for a in names:
for b in names:
ea = 1 / (1 + BASE**((elo_ratings[b] - elo_ratings[a]) / SCALE))
wins[a][b] = ea
wins[b][a] = 1 - ea
data = {
a: [wins[a][b] if a != b else np.NAN for b in names]
for a in names
}
df = pd.DataFrame(data, index=names)
df.index.name = 'model_a'
df.columns.name = 'model_b'
return df.T
def get_win_rate_column(df, column, baseline='gpt4-0314'):
to_dict = df[['model', column]].set_index('model').to_dict()[column]
win_rate_table = predict_win_rate(to_dict)
return win_rate_table[baseline].fillna(0.5).apply(
lambda x: round(x * 100, 2))
@DICT_POSTPROCESSORS.register_module('arenahard')
def arenahard_postprocess(
output: dict,
output_path: str,
) -> dict:
judged_answers, references = get_judgeanswer_and_reference(
output, output_path, post_process_arenahard)
if len(judged_answers) == 0:
scores = None
battles = get_battles_from_judgment(
judged_answers,
references,
)
bootstrap_online_elo = compute_mle_elo(battles)
np.random.seed(42)
bootstrap_elo_lu = get_bootstrap_result(battles, compute_mle_elo, 100)
stats = pd.DataFrame()
stats['results'] = None
stats['results'] = stats['results'].astype('object')
for i, model in enumerate(bootstrap_online_elo.index):
assert model in bootstrap_elo_lu.columns
stats.at[i, 'model'] = model
stats.at[i, 'score'] = bootstrap_online_elo[model]
# stats.at[i, 'lower'] = np.percentile(bootstrap_elo_lu[model], 2.5)
# stats.at[i, 'upper'] = np.percentile(bootstrap_elo_lu[model], 97.5)
# stats.at[i, 'results'] = bootstrap_elo_lu[model].tolist()
stats['score'] = get_win_rate_column(stats, 'score', 'gpt4-0314').tolist()
models = stats['model']
scores = stats['score']
if models[0] == 'gpt4-0314':
score = scores[1]
else:
score = scores[0]
results = {'score': score}
results['details'] = output
return results
@DICT_POSTPROCESSORS.register_module('arenahard_bradleyterry')
def arenahard_bradleyterry_postprocess(
output: dict,
output_path: str,
) -> dict:
judged_answers, references = get_judgeanswer_and_reference(
result=output,
filename=output_path,
post_process=post_process_arenahard,
)
if 'prediction1' not in references[0]:
raise ValueError(
'prediction1 not in references. Set `keep_predictions=True` for LMEvaluator in dataset config and retry.'
)
if 'prediction2' not in references[0]:
raise ValueError(
'prediction2 not in references. Set `keep_predictions=True` for LMEvaluator in dataset config and retry.'
)
results = {}
matches = []
for judged_answer, reference in zip(judged_answers, references):
cur_dict = {}
if judged_answer in ['A>>B', 'B<<A', 'A>B', 'B<A']:
cur_dict['winner'] = 'model_a'
elif judged_answer in ['A=B', 'B=A']:
cur_dict['winner'] = 'tie'
elif judged_answer in ['A<B', 'B>A', 'A<<B', 'B>>A']:
cur_dict['winner'] = 'model_b'
else:
continue
cur_dict['capability'] = reference['capability']
cur_dict['model_a'] = reference['answer1']
cur_dict['model_b'] = reference['answer2']
cur_dict['prediction1'] = reference['prediction1']
cur_dict['prediction2'] = reference['prediction2']
matches.append(cur_dict)
### ---------- Add Style Metadata ---------- ###
matches = get_element_counts(
data=matches,
column='prediction1',
suffix='_a',
)
matches = get_element_counts(
data=matches,
column='prediction2',
suffix='_b',
)
results['matches'] = matches
# results["details"] = output
return results