Spaces:
Running
Running
Upload app.py with huggingface_hub
Browse files
app.py
CHANGED
@@ -116,8 +116,7 @@ def expand_query(state):
|
|
116 |
Dict: The updated state with the expanded query.
|
117 |
"""
|
118 |
print("---------Expanding Query---------")
|
119 |
-
system_message =
|
120 |
-
'''
|
121 |
You are a domain expert assisting in answering questions related to research papers.
|
122 |
Convert the user query into something that a nutritionist would understand. Use domain related words.
|
123 |
Return 3 related search queries based on the user's request seperated by newline.
|
@@ -200,8 +199,7 @@ def craft_response(state: Dict) -> Dict:
|
|
200 |
Dict: The updated state with the generated response.
|
201 |
"""
|
202 |
print("---------craft_response---------")
|
203 |
-
system_message =
|
204 |
-
'''
|
205 |
Generates a response to a user query and context provided.
|
206 |
|
207 |
Parameters:
|
@@ -250,10 +248,9 @@ def score_groundedness(state: Dict) -> Dict:
|
|
250 |
Dict: The updated state with the groundedness score.
|
251 |
"""
|
252 |
print("---------check_groundedness---------")
|
253 |
-
system_message =
|
254 |
-
'''
|
255 |
You are tasked with rating AI generated answers to questions posed by users.
|
256 |
-
Please act as an impartial judge and evaluate the quality of the provided answer which attempts to answer the provided question based on a provided context.
|
257 |
In the input, the context is {context}, while the AI generated response is {response}.
|
258 |
|
259 |
Evaluation criteria:
|
@@ -300,11 +297,10 @@ def check_precision(state: Dict) -> Dict:
|
|
300 |
Dict: The updated state with the precision score.
|
301 |
"""
|
302 |
print("---------check_precision---------")
|
303 |
-
system_message =
|
304 |
-
'''
|
305 |
Given question, answer and context verify if the context was useful in arriving at the given answer.
|
306 |
Give verdict as "1" if useful and "0" if not useful.
|
307 |
-
Output your result as a float number between 0 and 1
|
308 |
Give verdict as a scaled numeric value of type float between 0 and 1, such that
|
309 |
0 or near 0 if it is least useful, 0.5 or near 0.5 if retry is warranted, and 1 or close to 1 is most useful.
|
310 |
Do not show any instructions for deriving your answer.
|
@@ -338,9 +334,8 @@ def refine_response(state: Dict) -> Dict:
|
|
338 |
"""
|
339 |
print("---------refine_response---------")
|
340 |
|
341 |
-
system_message =
|
342 |
-
|
343 |
-
Since the last response failded the groundedness test, and is deemed not satisfactory,
|
344 |
use the feedback in terms of the query, context and the last response
|
345 |
to identify potential gaps, ambiguities, or missing details, and
|
346 |
to suggest improvements to enhance accuracy and completeness of the response.
|
@@ -374,9 +369,8 @@ def refine_query(state: Dict) -> Dict:
|
|
374 |
Dict: The updated state with query refinement suggestions.
|
375 |
"""
|
376 |
print("---------refine_query---------")
|
377 |
-
system_message =
|
378 |
-
|
379 |
-
Since the last response failded the precision test, and is deemed not satisfactory,
|
380 |
use the feedback in terms of the query, context and re-generate extended queries
|
381 |
to identify specific keywords, scope refinements, or missing details, and
|
382 |
to provides structured suggestions for improvement to enhance accuracy and completeness of the response.
|
|
|
116 |
Dict: The updated state with the expanded query.
|
117 |
"""
|
118 |
print("---------Expanding Query---------")
|
119 |
+
system_message = '''
|
|
|
120 |
You are a domain expert assisting in answering questions related to research papers.
|
121 |
Convert the user query into something that a nutritionist would understand. Use domain related words.
|
122 |
Return 3 related search queries based on the user's request seperated by newline.
|
|
|
199 |
Dict: The updated state with the generated response.
|
200 |
"""
|
201 |
print("---------craft_response---------")
|
202 |
+
system_message = '''
|
|
|
203 |
Generates a response to a user query and context provided.
|
204 |
|
205 |
Parameters:
|
|
|
248 |
Dict: The updated state with the groundedness score.
|
249 |
"""
|
250 |
print("---------check_groundedness---------")
|
251 |
+
system_message = '''
|
|
|
252 |
You are tasked with rating AI generated answers to questions posed by users.
|
253 |
+
Please act as an impartial judge and evaluate the quality of the provided answer which attempts to answer the provided question based on a provided context.
|
254 |
In the input, the context is {context}, while the AI generated response is {response}.
|
255 |
|
256 |
Evaluation criteria:
|
|
|
297 |
Dict: The updated state with the precision score.
|
298 |
"""
|
299 |
print("---------check_precision---------")
|
300 |
+
system_message = '''
|
|
|
301 |
Given question, answer and context verify if the context was useful in arriving at the given answer.
|
302 |
Give verdict as "1" if useful and "0" if not useful.
|
303 |
+
Output your result as a float number between 0 and 1
|
304 |
Give verdict as a scaled numeric value of type float between 0 and 1, such that
|
305 |
0 or near 0 if it is least useful, 0.5 or near 0.5 if retry is warranted, and 1 or close to 1 is most useful.
|
306 |
Do not show any instructions for deriving your answer.
|
|
|
334 |
"""
|
335 |
print("---------refine_response---------")
|
336 |
|
337 |
+
system_message = '''
|
338 |
+
Since the last response failded the groundedness test, and is deemed not satisfactory,
|
|
|
339 |
use the feedback in terms of the query, context and the last response
|
340 |
to identify potential gaps, ambiguities, or missing details, and
|
341 |
to suggest improvements to enhance accuracy and completeness of the response.
|
|
|
369 |
Dict: The updated state with query refinement suggestions.
|
370 |
"""
|
371 |
print("---------refine_query---------")
|
372 |
+
system_message = '''
|
373 |
+
Since the last response failded the precision test, and is deemed not satisfactory,
|
|
|
374 |
use the feedback in terms of the query, context and re-generate extended queries
|
375 |
to identify specific keywords, scope refinements, or missing details, and
|
376 |
to provides structured suggestions for improvement to enhance accuracy and completeness of the response.
|