Spaces:
Running
Running
File size: 7,258 Bytes
81108ac |
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 |
"""
This program helps us explore model's responses to the benchmark. It is a web
app that displays the following:
1. A list of benchmark items loaded from puzzles_cleaned.csv. The list shows
the columns ID, challenge, and answer.
2. When we select a puzzle from the list, we see the transcript, Explanation,
and Editor's Note in textboxes. (Scrollable since they can be long.)
3. The list in (1) also has a column for each model, with checkboxes indicating
whether the model's response is correct or not. We load the model responses
from results.duckdb. That file has a table called completions with
columns 'prompt_id', 'parent_dir', and 'completion'. The prompt_id can be
joined with ID from puzzles_cleaned.csv. The parent_dir is the model name.
The completion is the model response, which we compare with the answer from
puzzles_cleaned.csv using the function check_answer defined below.
4. Finally, when an item is selected from the list, we get a dropdown that lets
us select a model to see the completion from that model.
Note that not every model has a response for every puzzle.
"""
import re
import duckdb
import gradio as gr
import textwrap
def split_into_words(text: str) -> list:
return re.findall(r'\b\w+\b', text.lower())
def all_words_match(completion: str, answer: str) -> bool:
answer_words = split_into_words(answer)
completion = completion.lower()
return all(word in completion for word in answer_words)
def answer_without_thoughts(completion: str) -> str:
if "<think>" not in completion[:200]:
return completion
chunks = completion.split("</think>")
if len(chunks) <= 1:
return ""
return chunks[-1].strip()
def check_answer(completion: str, answer: str) -> bool:
"""
Check if all words in the answer are in the completion, in the same order.
"""
completion_words = split_into_words(answer_without_thoughts(completion))
answer_words = split_into_words(answer)
indices = []
for word in answer_words:
if word in completion_words:
indices.append(completion_words.index(word))
else:
return False
return indices == sorted(indices) or indices == sorted(indices, reverse=True)
def clip_text(text: str, width: int) -> str:
return text if len(text) <= width else text[:width] + "..."
def wrap_text(text: str, width: int) -> str:
return textwrap.fill(text, width=width)
def get_model_response(prompt_id, model_name):
query = f"""
SELECT completion FROM results.completions
WHERE prompt_id = {prompt_id} AND parent_dir = '{model_name}'
"""
response = conn.sql(query).fetchone()
return response[0] if response else None
def display_puzzle(puzzle_id):
query = f"""
SELECT challenge, answer, transcript, Explanation, "Editor's Notes"
FROM challenges
WHERE ID = {puzzle_id}
"""
puzzle = conn.sql(query).fetchone()
return puzzle if puzzle else (None, None,None, None, None)
def display_model_response(puzzle_id, model_name):
response = get_model_response(puzzle_id, model_name)
split_thoughts = response.split("</think>")
if len(split_thoughts) > 1:
response = split_thoughts[-1].strip()
return "From " + model_name + ":\n" + response if response else "No response from this model."
conn = duckdb.connect(":memory:")
conn.execute("ATTACH DATABASE 'results.duckdb' AS results")
conn.execute("CREATE TABLE challenges as SELECT * FROM 'puzzles_cleaned.csv'")
conn.create_function("check_answer", check_answer)
conn.create_function("clip_text", clip_text)
conn.create_function("wrap_text", wrap_text)
# Get all unique model names
model_names = [item[0] for item in conn.sql("SELECT DISTINCT parent_dir FROM results.completions").fetchall()]
# Just for display.
cleaned_model_names = [name.replace("completions-", "") for name in model_names]
print(cleaned_model_names)
def build_table():
# Construct the query to create two columns for each model: MODEL_answer and MODEL_ok
query = """
SELECT c.ID, c.challenge, wrap_text(c.answer, 40) AS answer,
"""
model_correct_columns = []
for model in model_names:
normalized_model_name = model.replace("-", "_")
model_correct_columns.append(normalized_model_name + "_ok")
query += f"""
MAX(CASE WHEN r.parent_dir = '{model}' THEN r.completion ELSE NULL END) AS {normalized_model_name}_answer,
MAX(CASE WHEN r.parent_dir = '{model}' THEN check_answer(r.completion, c.answer) ELSE NULL END) AS {normalized_model_name}_ok,
"""
query = query.rstrip(',') # Remove the trailing comma
query += """
clip_text(c.challenge, 40) as challenge_clipped,
FROM challenges c
LEFT JOIN results.completions r
ON c.ID = r.prompt_id
GROUP BY c.ID, c.challenge, c.answer
"""
joined_df = conn.sql(query).fetchdf()
# Transform the model_correct columns to use emojis
for model in model_names:
normalized_model_name = model.replace("-", "_")
joined_df[normalized_model_name + '_ok'] = joined_df[normalized_model_name + '_ok'].apply(
lambda x: "✅" if x == 1 else ("❌" if x == 0 else "❓")
)
return joined_df, model_correct_columns
joined_df, model_correct_columns = build_table()
relabelled_df = joined_df[['ID', 'challenge_clipped', 'answer', *model_correct_columns]].rename(columns={
'ID': 'Puzzle ID',
'challenge_clipped': 'Challenge',
'answer': 'Answer',
**{model.replace("-", "_") + '_ok': model.replace("completions-", "") for model in model_names}
})
model_columns = {
index + 3: name for index, name in enumerate(model_names)
}
valid_model_indices = list(model_columns.keys())
default_model = model_columns[valid_model_indices[0]]
def create_interface():
with gr.Blocks() as demo:
# Using "markdown" as the datatype makes Gradio interpret newlines.
puzzle_list = gr.DataFrame(
value=relabelled_df,
datatype=["number", "str", "markdown", *["str"] * len(model_correct_columns)],
# headers=["ID", "Challenge", "Answer", *cleaned_model_names],
)
model_response = gr.Textbox(label="Model Response", interactive=False)
challenge = gr.Textbox(label="Challenge", interactive=False)
answer = gr.Textbox(label="Answer", interactive=False)
explanation = gr.Textbox(label="Explanation", interactive=False)
editors_note = gr.Textbox(label="Editor's Note", interactive=False)
transcript = gr.Textbox(label="Transcript", interactive=False)
def update_puzzle(evt: gr.SelectData):
row = evt.index[0]
model_index = evt.index[1]
model_name = model_columns[model_index] if model_index in valid_model_indices else default_model
return (*display_puzzle(row), display_model_response(row, model_name))
puzzle_list.select(
fn=update_puzzle,
inputs=[],
outputs=[challenge, answer, transcript, explanation, editors_note, model_response]
)
demo.launch()
if __name__ == "__main__":
create_interface()
|