BtB-ExpC commited on
Commit
f0c59b9
·
1 Parent(s): 4d55c1b

added the cap @ 16 learning objectives

Browse files
config/llm_config.py CHANGED
@@ -44,7 +44,8 @@ llms = {
44
  "GPT-4 Turbo (low temp)": create_openai_llm("gpt-4-turbo-2024-04-09", LOW),
45
 
46
  # OpenAI reasoning models (no temperature)
47
- "o1": create_openai_reasoning_llm("o1-2024-12-17", reasoning_effort="high"),
 
48
  "o3-mini (low reasoning_effort)": create_openai_reasoning_llm("o3-mini", reasoning_effort="low"),
49
  "o3-mini (medium reasoning_effort)": create_openai_reasoning_llm("o3-mini", reasoning_effort="medium"),
50
  "o3-mini (high reasoning_effort)": create_openai_reasoning_llm("o3-mini", reasoning_effort="high"),
 
44
  "GPT-4 Turbo (low temp)": create_openai_llm("gpt-4-turbo-2024-04-09", LOW),
45
 
46
  # OpenAI reasoning models (no temperature)
47
+ "o1 (low reasoning_effort)": create_openai_reasoning_llm("o1-2024-12-17", reasoning_effort="low"),
48
+ "o1 (high reasoning_effort)": create_openai_reasoning_llm("o1-2024-12-17", reasoning_effort="high"),
49
  "o3-mini (low reasoning_effort)": create_openai_reasoning_llm("o3-mini", reasoning_effort="low"),
50
  "o3-mini (medium reasoning_effort)": create_openai_reasoning_llm("o3-mini", reasoning_effort="medium"),
51
  "o3-mini (high reasoning_effort)": create_openai_reasoning_llm("o3-mini", reasoning_effort="high"),
config/system_prompt_texts.py CHANGED
@@ -273,10 +273,10 @@ Before extracting learning objectives:
273
  </examples>
274
 
275
  ### Content Quality
276
- * **Self-contained**: Must be understandable without relying on any outside context
277
- * **Falsifiable**: Must be unambiguously, demonstrably true or false
278
  * **Factually Equivalent**: Must represent exactly the knowledge as written
279
- * **Specific**: Must express the smallest coherent, testable knowledge unit, instead of several things at once
280
 
281
  <examples>
282
  <self_contained_example>
@@ -299,7 +299,9 @@ Before extracting learning objectives:
299
  </examples>
300
 
301
  ### Additional notes
302
- *
 
 
303
 
304
  ### Language Precision
305
  * Avoid universal terms ("always", "never") unless 100% accurate, and there are in fact no exceptions
@@ -308,7 +310,7 @@ Before extracting learning objectives:
308
  * Use "important" only when there is no other option to say something more meaningful ("X is important" doesn't say much)
309
 
310
  ## Quality Assurance Process
311
- 1. Extract all potential learning objectives
312
  2. For each objective, verify it meets ALL requirements
313
  3. Refine and potentially split objectives until each one is:
314
  - Maximally specific
@@ -337,7 +339,7 @@ Before extracting learning objectives:
337
  - De student weet dat je dagelijks oefent om zo objectief (zonder je eigen mening) mogelijk te observeren.
338
  - De student weet dat je aan de zorgvrager vraagt of jouw observatie klopt.
339
  </better>
340
- <explanation>Now they are stated mirroring the language of the text, without introducing new, potentially difficult words.</explanation>
341
  </language_level_examples>
342
  </examples>
343
 
 
273
  </examples>
274
 
275
  ### Content Quality
276
+ * **Self-contained**: Must be understandable without relying on any outside context, like 'previous' learning objectives (they will in fact be presented to the student non-chronologically, so cannot build on each other)
277
+ * **Falsifiable**: Must be unambiguously, demonstrably true or false
278
  * **Factually Equivalent**: Must represent exactly the knowledge as written
279
+ * **Specific**: Must express the smallest coherent, testable knowledge unit, instead of several things at once. For composite knowledge, like
280
 
281
  <examples>
282
  <self_contained_example>
 
299
  </examples>
300
 
301
  ### Additional notes
302
+ * If a term is explained in the text, and it's central to the topic, knowing what it means should also become a learning objective
303
+ * If a term is explained in the text, also explain it briefly (between parentheses) in each learning objective that uses it
304
+ * If the text is so knowledge dense that it contains more than 16 potential learning objectives, cap them at around that number and eliminate the less important ones. For non-hierarchical learning objectives (all equally important), just trim randomly
305
 
306
  ### Language Precision
307
  * Avoid universal terms ("always", "never") unless 100% accurate, and there are in fact no exceptions
 
310
  * Use "important" only when there is no other option to say something more meaningful ("X is important" doesn't say much)
311
 
312
  ## Quality Assurance Process
313
+ 1. Extract all potential learning objectives (up to ~16 max)
314
  2. For each objective, verify it meets ALL requirements
315
  3. Refine and potentially split objectives until each one is:
316
  - Maximally specific
 
339
  - De student weet dat je dagelijks oefent om zo objectief (zonder je eigen mening) mogelijk te observeren.
340
  - De student weet dat je aan de zorgvrager vraagt of jouw observatie klopt.
341
  </better>
342
+ <explanation>Now they are stated mirroring the language of the text, without introducing new, potentially difficult words. Words that are explained in the text, need to be explained in each learning objective that reuses it between parentheses as well.</explanation>
343
  </language_level_examples>
344
  </examples>
345
 
test samples.md CHANGED
@@ -92,10 +92,13 @@ Het hart bestaat uit vier holtes: twee boezems aan de bovenkant en twee kamers a
92
  - De student weet dat de boezems van het hart aan de bovenkant zitten.
93
  - De student weet dat het hart uit twee kamers bestaat.
94
  - De student weet dat de kamers van het hart aan de onderkant zitten.
95
- ## 2 Require "Theorie"
96
  - De student weet dat als je een onderbuikgevoel krijgt, je hierop moet letten, moet onderzoeken waar het vandaan komt en het met collega’s moet bespreken.
97
- ## 3 Must be tested approximately (not with 1 year precision)
 
98
  - De student weet dat de Wet Bopz sinds 1994 in gebruik was.
 
 
99
 
100
 
101
 
 
92
  - De student weet dat de boezems van het hart aan de bovenkant zitten.
93
  - De student weet dat het hart uit twee kamers bestaat.
94
  - De student weet dat de kamers van het hart aan de onderkant zitten.
95
+ ## Require "Theorie"
96
  - De student weet dat als je een onderbuikgevoel krijgt, je hierop moet letten, moet onderzoeken waar het vandaan komt en het met collega’s moet bespreken.
97
+ - De student weet dat je dagelijks oefent om zo objectief (zonder je eigen mening) mogelijk te observeren.
98
+ ## Must be tested approximately (not with 1 year precision)
99
  - De student weet dat de Wet Bopz sinds 1994 in gebruik was.
100
+ ## Require Extra info (explanation)
101
+ - De student weet dat de Wet Bopz de rechten en plichten van onvrijwillig opgenomen zorgvragers regelde, zoals het behandelplan, verlof en juridische bijstand.
102
 
103
 
104