-
Notifications
You must be signed in to change notification settings - Fork 40
/
tools.py
206 lines (176 loc) · 9.37 KB
/
tools.py
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
import os
import openai
from utils.references import References
from utils.gpt_interaction import GPTModel
from prompts import SYSTEM
from langchain.tools import BaseTool
from pydantic import BaseModel, Field
from typing import Optional, Type
MAX_TOKENS = 2048
openai.api_key = os.getenv("OPENAI_API_KEY")
default_model = os.getenv("DEFAULT_MODEL")
if default_model is None:
default_model = "gpt-3.5-turbo-16k"
llm = GPTModel(model=default_model, delay=1)
paper_system_prompt = '''You are an assistant designed to propose choices of research direction.
The user will input questions or some keywords of a fields. You need to generate some paper titles and main contributions. Ensure follow the following instructions:
Instruction:
- Your response should follow the JSON format.
- Your response should have the following structure:
{
"your suggested paper title":
{
"summary": "an overview introducing what this paper will include",
"contributions": {
"contribution1": {"statement": "briefly describe this contribution", "reason": "reason why this contribution can make this paper outstanding"},
"contribution2": {"statement": "briefly describe this contribution", "reason": "reason why this contribution can make this paper outstanding"},
...
}
}
"your suggested paper title":
{
"summary": "an overview introducing what this paper will include",
"contributions": {
"contribution1": {"statement": "briefly describe this contribution", "reason": "reason why this contribution can make this paper outstanding"},
"contribution2": {"statement": "briefly describe this contribution", "reason": "reason why this contribution can make this paper outstanding"},
...
}
}
...
}
- Please list three to five suggested title and at least three contributions for each paper.
'''
contribution_system_prompt = '''You are an assistant designed to criticize the contributions of a paper. You will be provided Paper's Title, References and Contributions. Ensure follow the following instructions:
Instruction:
- Your response should follow the JSON format.
- Your response should have the following structure:
{
"title": "the title provided by the user",
"comment": "your thoughts on if this title clearly reflects the key ideas of this paper and explain why"
"contributions": {
"contribution1": {"statement": "briefly describe what the contribution is",
"reason": "reason why the user claims it is a contribution",
"judge": "your thought about if this is a novel contribution and explain why",
"suggestion": "your suggestion on how to modify the research direction to enhance the novelty "},
"contribution2": {"statement": "briefly describe what the contribution is",
"reason": "reason why the user claims it is a contribution",
"judge": "your thought about if this is a novel contribution and explain why",
"suggestion": "your suggestion on how to modify the research direction to enhance the novelty "},
...
}
}
- You need to carefully check if the claimed contribution has been made in the provided references, which makes the contribution not novel.
- You also need to propose your concerns on if any of contributions could be incremental or just a mild modification on an existing work.
'''
def find_research_directions(research_field):
output, _ = llm(systems=paper_system_prompt, prompts=research_field, return_json=False)
return output
def find_references(title, contributions):
max_tokens = MAX_TOKENS
ref = References(title=title, description=f"{contributions}")
keywords, _ = llm(systems=SYSTEM["keywords"], prompts=title, return_json=True)
keywords = {keyword: 10 for keyword in keywords}
ref.collect_papers(keywords)
ref_prompt = ref.to_prompts(max_tokens=max_tokens)
return ref_prompt
def judge_novelty(title, contributions):
max_tokens = MAX_TOKENS
ref = References(title=title, description=f"{contributions}")
keywords, _ = llm(systems=SYSTEM["keywords"], prompts=title, return_json=True)
keywords = {keyword: 10 for keyword in keywords}
ref.collect_papers(keywords)
ref_prompt = ref.to_prompts(max_tokens=max_tokens)
prompt = f"Title: {title}\n References: {ref_prompt}\n Contributions: {contributions}"
output, _ = llm(systems=contribution_system_prompt, prompts=prompt, return_json=False)
return output
functions = [
{
"name": "find_research_directions",
"description": "when your student has already shown interests in a specific topic and provided a rough description of potential contributions, help your student to dive this direction deeper",
"parameters": {
"type": "object",
"properties": {
"research_description": {
"type": "string",
"description": "a paragraph with details in English describing "
"(1) what is the main problem you are trying to solve "
"(2) what is the main novelty of this idea (3) how to complete this research."
}
},
"required": ["research_description"],
},
},
{
"name": "find_references",
"description": "find references for given details of a paper",
"parameters": {
"type": "object",
"properties": {
"title": {
"type": "string",
"description": "the title (in English) of the academic paper your student will write.",
},
"contributions": {"type": "string",
"description": "a general description on the contributions of this paper in English."
"If there are multiple contributions, index them with numbers."},
},
"required": ["title", "contributions"],
},
},
{
"name": "judge_novelty",
"description": "evaluate the novelty of a paper given its title and main contributions",
"parameters": {
"type": "object",
"properties": {
"title": {
"type": "string",
"description": "the title (in English) of the academic paper your student will write.",
},
"contributions": {"type": "string",
"description": "a general description on the contributions of this paper in English."
"If there are multiple contributions, index them with numbers."},
},
"required": ["title", "contributions"],
},
}
]
TOOLS = {"find_research_directions": find_research_directions, "find_references": find_references, "judge_novelty": judge_novelty}
class FindResearchDirectionsCheckInput(BaseModel):
research_description: str = Field(..., description="a paragraph with details in English describing (1) what is the main problem you are trying to solve "
"(2) what is the main novelty of this idea (3) how to complete this research.")
class TitleDescriptionCheckInput(BaseModel):
title: str = Field(..., description="the title of the academic paper your student will write in English.")
contributions: str = Field(..., description="a general description on the contributions of this paper in English."
"If there are multiple contributions, index them with numbers.")
class FindResearchDirectionsTool(BaseTool):
name = "find_research_directions"
description = """Useful when your student has already shown interests in a specific topic and provided a rough description of
potential contributions and you need to help your student to dive this direction deeper for your student.
"""
def _run(self, research_description: str):
response = find_research_directions(research_description)
return response
def _arun(self, research_field: str):
raise NotImplementedError("This tool does not support async")
args_schema: Optional[Type[BaseModel]] = FindResearchDirectionsCheckInput
class JudgeNoveltyTool(BaseTool):
name = "judge_novelty"
description = """Useful when you need to evaluate the novelty of your student's idea.
"""
def _run(self, title: str, contributions: str):
response = judge_novelty(title, contributions)
return response
def _arun(self, title: str, contributions: str):
raise NotImplementedError("This tool does not support async")
args_schema: Optional[Type[BaseModel]] = TitleDescriptionCheckInput
class FindReferencesTool(BaseTool):
name = "find_references"
description = """Useful when you need to find references for a paper.
"""
def _run(self, title: str, contributions: str):
response = find_references(title, contributions)
return response
def _arun(self, title: str, contributions: str):
raise NotImplementedError("This tool does not support async")
args_schema: Optional[Type[BaseModel]] = TitleDescriptionCheckInput