Improvements (#20)

- set intersection: simplified merge prompt
- sorting: fixed split prompt
- fixed white space issues
- fixed typos
- fixed a few consistency issues

examples/doc_merge/doc_merge.py

@@ -65,7 +65,7 @@ Here is the summary NDA <S>:
 """
 
     score_prompt_base = """The following NDA <S> merges NDAs <Doc1> - <Doc{num}>.
-Please score the merged NDA <S> in terms of how much redundant information is contained, independent of the original NDAs, as well as how much information if retained from the original NDAs. 
+Please score the merged NDA <S> in terms of how much redundant information is contained, independent of the original NDAs, as well as how much information is retained from the original NDAs.
 A score of 10 for redundancy implies that absolutely no information is redundant, while a score of 0 implies that at least half of the information is redundant (so everything is at least mentioned twice).
 A score of 10 for retained information implies that all information from the original NDAs is retained, while a score of 0 implies that no information is retained.
 You may provide reasoning for your scoring, but the final score for redundancy should be between the tags <Redundancy> and </Redundancy>, and the final score for retained information should be between the tags <Retained> and </Retained>, without any additional text within any of those tags.

examples/doc_merge/example_prompts_document_merging.md

@@ -27,7 +27,7 @@ Here are NDAs <Doc1> - <Doc4>
 Replace `{doc1}`, `{doc2}`, `{doc3}`, and `{doc4}` with the NDAs that are supposed to be merged and `{s}` with the merged summary NDA that is to be scored.
 ```
 The following NDA <S> merges NDAs <Doc1> - <Doc4>.
-Please score the merged NDA <S> in terms of how much redundant information is contained, independent of the original NDAs, as well as how much information if retained from the original NDAs. 
+Please score the merged NDA <S> in terms of how much redundant information is contained, independent of the original NDAs, as well as how much information is retained from the original NDAs.
 A score of 10 for redundancy implies that absolutely no information is redundant, while a score of 0 implies that at least half of the information is redundant (so everything is at least mentioned twice).
 A score of 10 for retained information implies that all information from the original NDAs is retained, while a score of 0 implies that no information is retained.
 You may provide reasoning for your scoring, but the final score for redundancy should be between the tags <Redundancy> and </Redundancy>, and the final score for retained information should be between the tags <Retained> and </Retained>, without any additional text within any of those tags.
@@ -131,10 +131,10 @@ NON-DISCLOSURE AGREEMENT (NDA)
 
 1. Agreement between [Your Company Name] and [Recipient Name] on [Date].
 2. Information sharing for the purpose of [specific project or purpose].
-3. ”ConfidentialInformation”includesallpotentiallycommerciallyvaluableinformation,specificallysoftwaredevelopment tactics, processes, and in-house research results.
+3. "ConfidentialInformation" includes all potentially commercially valuable information, specifically software development tactics, processes, and in-house research results.
 4. Receiving party is obligated to protect the Confidential Information, use it solely for the disclosed purpose, and not disclose it without consent.
 5. Breach penalties include injunctive relief, other remedies, and a $200,000 fee per breach.
-6. The Agreement applies to the Parties and their successors and assigns. It contains all related agreements and lack of enforcement doesn’t imply waiver.
+6. The Agreement applies to the Parties and their successors and assigns. It contains all related agreements and lack of enforcement doesn't imply waiver.
 7. The Agreement is under the laws of [State].
 8. Signed by [Your Company Name] and [Recipient Name] at the above date.
 ```
@@ -215,7 +215,6 @@ SECRECY & DISCLOSURE AGREEMENT
 
 Categorization of Sensitive Data:
 Includes:
-
 a. Internal software blueprints.
 b. Intellectual property awaiting patents.
 
@@ -331,7 +330,6 @@ SECRECY & DISCLOSURE AGREEMENT
 
 Categorization of Sensitive Data:
 Includes:
-
 a. Internal software blueprints.
 b. Intellectual property awaiting patents.
 
@@ -776,7 +774,7 @@ Confirmed and endorsed on [Date].
 ### Step 1 Scoring for Best Response - Prompt
 ```
 The following NDA <S> merges NDAs <Doc1> - <Doc4>.
-Please score the merged NDA <S> in terms of how much redundant information is contained, independent of the original NDAs, as well as how much information if retained from the original NDAs. 
+Please score the merged NDA <S> in terms of how much redundant information is contained, independent of the original NDAs, as well as how much information is retained from the original NDAs.
 A score of 10 for redundancy implies that absolutely no information is redundant, while a score of 0 implies that at least half of the information is redundant (so everything is at least mentioned twice).
 A score of 10 for retained information implies that all information from the original NDAs is retained, while a score of 0 implies that no information is retained.
 You may provide reasoning for your scoring, but the final score for redundancy should be between the tags <Redundancy> and </Redundancy>, and the final score for retained information should be between the tags <Retained> and </Retained>, without any additional text within any of those tags.
@@ -835,7 +833,6 @@ CONFIDENTIALITY & NON-DISCLOSURE AGREEMENT
 
 Specification of Protected Information:
 Particularly:
-
 a. System designs and architectural schematics.
 b. Proprietary computational algorithms.
 
@@ -870,7 +867,6 @@ SECRECY & DISCLOSURE AGREEMENT
 
 Categorization of Sensitive Data:
 Includes:
-
 a. Internal software blueprints.
 b. Intellectual property awaiting patents.
 
@@ -1067,7 +1063,6 @@ CONFIDENTIALITY & NON-DISCLOSURE AGREEMENT
 
 Specification of Protected Information:
 Particularly:
-
 a. System designs and architectural schematics.
 b. Proprietary computational algorithms.
 
@@ -1102,7 +1097,6 @@ SECRECY & DISCLOSURE AGREEMENT
 
 Categorization of Sensitive Data:
 Includes:
-
 a. Internal software blueprints.
 b. Intellectual property awaiting patents.
 
@@ -1803,7 +1797,6 @@ CONFIDENTIALITY & NON-DISCLOSURE AGREEMENT
 
 Specification of Protected Information:
 Particularly:
-
 a. System designs and architectural schematics.
 b. Proprietary computational algorithms.
 
@@ -1838,7 +1831,6 @@ SECRECY & DISCLOSURE AGREEMENT
 
 Categorization of Sensitive Data:
 Includes:
-
 a. Internal software blueprints.
 b. Intellectual property awaiting patents.
 
@@ -2441,7 +2433,7 @@ Confirmed and endorsed on [Date].
 
 ```
 The following NDA <S> merges NDAs <Doc1> - <Doc4>.
-Please score the merged NDA <S> in terms of how much redundant information is contained, independent of the original NDAs, as well as how much information if retained from the original NDAs. 
+Please score the merged NDA <S> in terms of how much redundant information is contained, independent of the original NDAs, as well as how much information is retained from the original NDAs.
 A score of 10 for redundancy implies that absolutely no information is redundant, while a score of 0 implies that at least half of the information is redundant (so everything is at least mentioned twice).
 A score of 10 for retained information implies that all information from the original NDAs is retained, while a score of 0 implies that no information is retained.
 You may provide reasoning for your scoring, but the final score for redundancy should be between the tags <Redundancy> and </Redundancy>, and the final score for retained information should be between the tags <Retained> and </Retained>, without any additional text within any of those tags.
@@ -2500,7 +2492,6 @@ CONFIDENTIALITY & NON-DISCLOSURE AGREEMENT
 
 Specification of Protected Information:
 Particularly:
-
 a. System designs and architectural schematics.
 b. Proprietary computational algorithms.
 
@@ -2535,7 +2526,6 @@ SECRECY & DISCLOSURE AGREEMENT
 
 Categorization of Sensitive Data:
 Includes:
-
 a. Internal software blueprints.
 b. Intellectual property awaiting patents.
 

examples/keyword_counting/example_prompts_keyword_counting.md

@@ -178,11 +178,7 @@ Replace `{dictionary_1}` and `{dictionary_2}` with dictionaries of countries
 to be merged and `{dictionary_incorrect}` with incorrectly merged dictionary
 to be improved.
 ```
-<Instruction> The following 2 dictionaries were combined into the third dictionary below. 
+<Instruction> The following 2 dictionaries were combined into the third dictionary below. However, some mistakes occured and the third dictionary is incorrect. Please fix the third dictionary so that it contains the correct frequencies for each country. The correct frequencies are the sum of the frequencies from the first 2 dictionaries. If a country is not present in one of the dictionaries, add it to the final dictionary with the frequency from the other dictionary.
-However, some mistakes occured and the third dictionary is incorrect. 
-Please fix the third dictionary so that it contains the correct frequencies for each country. 
-The correct frequencies are the sum of the frequencies from the first 2 dictionaries. 
-If a country is not present in one of the dictionaries, add it to the final dictionary with the frequency from the other dictionary.
 </Instruction>
 
 <Example>
@@ -894,7 +890,8 @@ Combined Output:
 ```
 <Instruction> Combine the following 2 dictionaries, each containing the frequency of countries in a text, into a single dictionary.
 Simply add the frequencies together for each country and if a country is not present in one of the dictionaries, add it to the final dictionary with the frequency from the other dictionary.
-Only output the final merged dictionary without any additional text or thoughts! </Instruction>
+Only output the final merged dictionary without any additional text or thoughts!
+</Instruction>
 
 <Approach>
 To combine the 2 dictionaries into single one, follow these steps:
@@ -934,7 +931,8 @@ Combined Output:
 ```
 <Instruction> Combine the following 2 dictionaries, each containing the frequency of countries in a text, into a single dictionary.
 Simply add the frequencies together for each country and if a country is not present in one of the dictionaries, add it to the final dictionary with the frequency from the other dictionary.
-Only output the final merged dictionary without any additional text or thoughts! </Instruction>
+Only output the final merged dictionary without any additional text or thoughts!
+</Instruction>
 
 <Approach>
 To combine the 2 dictionaries into single one, follow these steps:

examples/set_intersection/example_prompts_set_intersection_032.md

@@ -45,7 +45,7 @@ Input Set 1: {set1}
 Input Set 2: {set2}
 ```
 
-### Aggregate: merge_prompt
+### AGGREGATE: merge_prompt
 Replace `{input1}` and `{input2}` with the lists/sets to be merged.
 ```
 <Instruction> Merge the following 2 lists of length 16 each, into one list of length 32 by appending the second list to the first list.

examples/set_intersection/set_intersection_032.py

@@ -147,7 +147,7 @@ Output:
 
 Input: {input}"""
 
-    got_merge_prompt = """<Instruction> Merge the following 2 lists of length {length1} each, into one list of length {length2} by appending the second list to the first list.
+    got_merge_prompt = """<Instruction> Merge the following 2 lists into one list by appending the second list to the first list.
 Only output the final list without any additional text or thoughts! </Instruction>
 
 List 1: {input1}
@@ -166,20 +166,10 @@ List 2: {input2}
         :raise AssertionError: If not exactly two thought states are provided.
         """
         assert len(state_dicts) == 2, "Expected two states for aggregation prompt."
-        len_input1 = len(utils.string_to_list(state_dicts[0]["current"]))
-        len_input2 = len(utils.string_to_list(state_dicts[1]["current"]))
-        if len_input1 == len_input2:
-            length = len_input1
-        elif len_input1 + len_input2 - 32 <= 16:
-            length = 16
-        else:
-            length = 32
 
         return self.got_merge_prompt.format(
             input1=state_dicts[0]["current"],
             input2=state_dicts[1]["current"],
-            length1=length,
-            length2=length * 2,
         )
 
     def generate_prompt(

examples/set_intersection/set_intersection_064.py

@@ -151,7 +151,7 @@ Output:
 
 Input: {input}"""
 
-    got_merge_prompt = """<Instruction> Merge the following 2 lists of length {length1} each, into one list of length {length2} by appending the second list to the first list.
+    got_merge_prompt = """<Instruction> Merge the following 2 lists into one list by appending the second list to the first list.
 Only output the final list without any additional text or thoughts! </Instruction>
 
 List 1: {input1}
@@ -171,20 +171,10 @@ List 2: {input2}
         """
 
         assert len(state_dicts) == 2, "Expected two states for aggregation prompt."
-        len_input1 = len(utils.string_to_list(state_dicts[0]["current"]))
-        len_input2 = len(utils.string_to_list(state_dicts[1]["current"]))
-        if len_input1 == len_input2:
-            length = len_input1
-        elif len_input1 + len_input2 - 32 <= 16:
-            length = 16
-        else:
-            length = 32
 
         return self.got_merge_prompt.format(
             input1=state_dicts[0]["current"],
             input2=state_dicts[1]["current"],
-            length1=length,
-            length2=length * 2,
         )
 
     def generate_prompt(

examples/set_intersection/set_intersection_128.py

@@ -159,7 +159,7 @@ Output:
 
 Input: {input}"""
 
-    got_merge_prompt = """<Instruction> Merge the following 2 lists of length {length1} each, into one list of length {length2} by appending the second list to the first list.
+    got_merge_prompt = """<Instruction> Merge the following 2 lists into one list by appending the second list to the first list.
 Only output the final list without any additional text or thoughts! </Instruction>
 
 List 1: {input1}
@@ -178,22 +178,10 @@ List 2: {input2}
         :raise AssertionError: If not exactly two thought states are provided.
         """
         assert len(state_dicts) == 2, "Expected two states for aggregation prompt."
-        len_input1 = len(utils.string_to_list(state_dicts[0]["current"]))
-        len_input2 = len(utils.string_to_list(state_dicts[1]["current"]))
-        if len_input1 == len_input2:
-            length = len_input1
-        elif len_input1 + len_input2 - 32 <= 16:
-            length = 16
-        elif len_input1 + len_input2 - 64 <= 32:
-            length = 32
-        else:
-            length = 64
 
         return self.got_merge_prompt.format(
             input1=state_dicts[0]["current"],
             input2=state_dicts[1]["current"],
-            length1=length,
-            length2=length * 2,
         )
 
     def generate_prompt(

examples/sorting/example_prompts_sorting_032.md

@@ -4,19 +4,19 @@
 Replace `{input}` with the input list of numbers to be split.
 ```
 <Instruction> Split the following list of 32 numbers into 2 lists of 16 numbers each, the first list should contain the first 16 numbers and the second list the second 16 numbers.
-Only output the final 4 lists in the following format without any additional text or thoughts!:
+Only output the final 2 lists in the following format without any additional text or thoughts!:
-{
+{{
     "List 1": [3, 4, 3, 5, 7, 8, 1, ...],
     "List 2": [2, 9, 2, 4, 7, 1, 5, ...]
-} </Instruction>
+}} </Instruction>
 
 <Example>
 Input: [9, 6, 7, 7, 2, 0, 2, 2, 3, 5, 0, 9, 2, 2, 4, 4, 5, 2, 5, 1, 2, 8, 3, 8, 3, 9, 6, 0, 4, 2, 2, 3]
 Output:
-{
+{{
     "List 1": [9, 6, 7, 7, 2, 0, 2, 2, 3, 5, 0, 9, 2, 2, 4, 4],
     "List 2": [5, 2, 5, 1, 2, 8, 3, 8, 3, 9, 6, 0, 4, 2, 2, 3]
-}
+}}
 </Example>
 
 Input: {input}
@@ -74,7 +74,7 @@ Input: {input}
 Incorrectly Sorted: {incorrectly_sorted}
 ```
 
-### Aggregate: aggregate_prompt
+### AGGREGATE: aggregate_prompt
 Replace `{input_list1}` and `{input_list2}` with the input lists of numbers to be merged. `{length}` is the length of each input list and `{length_combined}` is the length of the combined list.
 ```
 <Instruction> Merge the following 2 sorted lists of length {length} each, into one sorted list of length {length_combined} using a merge sort style approach.
@@ -110,7 +110,7 @@ Merged list:
 ### Step 1 - Prompt
 ```
 <Instruction> Split the following list of 32 numbers into 2 lists of 16 numbers each, the first list should contain the first 16 numbers and the second list the second 16 numbers.
-Only output the final 4 lists in the following format without any additional text or thoughts!:
+Only output the final 2 lists in the following format without any additional text or thoughts!:
 {
     "List 1": [3, 4, 3, 5, 7, 8, 1, ...],
     "List 2": [2, 9, 2, 4, 7, 1, 5, ...]
@@ -196,7 +196,7 @@ Input: [0, 5, 6, 7, 1, 4, 5, 9, 4, 6, 2, 5, 8, 6, 2, 6]
 Only output the final merged list without any additional text or thoughts!:</Instruction>
 
 <Approach>
-To merge the two lists in a merge-sort style approach, foloow these steps:
+To merge the two lists in a merge-sort style approach, follow these steps:
 1. Compare the first element of both lists.
 2. Append the smaller element to the merged list and move to the next element in the list from which the smaller element came.
 3. Repeat steps 1 and 2 until one of the lists is empty.

examples/sorting/sorting_032.py

@@ -13,6 +13,7 @@ import json
 import csv
 from typing import Dict, List, Callable, Union
 from graph_of_thoughts import controller, language_models, operations, prompter, parser
+
 from . import utils
 
 
@@ -114,7 +115,7 @@ Incorrectly Sorted: {incorrectly_sorted}
 """
 
     got_split_prompt = """<Instruction> Split the following list of 32 numbers into 2 lists of 16 numbers each, the first list should contain the first 16 numbers and the second list the second 16 numbers.
-Only output the final 4 lists in the following format without any additional text or thoughts!:
+Only output the final 2 lists in the following format without any additional text or thoughts!:
 {{
     "List 1": [3, 4, 3, 5, 7, 8, 1, ...],
     "List 2": [2, 9, 2, 4, 7, 1, 5, ...]
@@ -135,7 +136,7 @@ Input: {input}"""
 Only output the final merged list without any additional text or thoughts!:</Instruction>
 
 <Approach>
-To merge the two lists in a merge-sort style approach, foloow these steps:
+To merge the two lists in a merge-sort style approach, follow these steps:
 1. Compare the first element of both lists.
 2. Append the smaller element to the merged list and move to the next element in the list from which the smaller element came.
 3. Repeat steps 1 and 2 until one of the lists is empty.

examples/sorting/sorting_064.py

@@ -140,7 +140,7 @@ Input: {input}"""
 Only output the final merged list without any additional text or thoughts!:</Instruction>
 
 <Approach>
-To merge the two lists in a merge-sort style approach, foloow these steps:
+To merge the two lists in a merge-sort style approach, follow these steps:
 1. Compare the first element of both lists.
 2. Append the smaller element to the merged list and move to the next element in the list from which the smaller element came.
 3. Repeat steps 1 and 2 until one of the lists is empty.

examples/sorting/sorting_128.py

@@ -148,7 +148,7 @@ Input: {input}"""
 Only output the final merged list without any additional text or thoughts!:</Instruction>
 
 <Approach>
-To merge the two lists in a merge-sort style approach, foloow these steps:
+To merge the two lists in a merge-sort style approach, follow these steps:
 1. Compare the first element of both lists.
 2. Append the smaller element to the merged list and move to the next element in the list from which the smaller element came.
 3. Repeat steps 1 and 2 until one of the lists is empty.