update docs

lm_eval/tasks/math_sat_cot.py

@@ -1,14 +1,9 @@
 """
-Solving Quantitative Reasoning Problems with Language Models
-https://arxiv.org/pdf/2206.14858.pdf
-
-Minerva CoT version of MMLU-STEM. See Appendix G for prompt reference. 
-
-`SUBJECTS_MCQA` consists of those subsets (STEM subsets that "use equations") for which the few-shot prompt is Listing 5 from the 
-Minerva Appendix G. SUBJECTS_CUSTOM use a special subject-specific prompt, which are listed as being in the supplementary material but 
-do not appear to be included in the current downloadable zip.
+SAT Math May 2023 questions that do not have figures.
 
+We use the version of the dataset found in the Huggingface dataset `mcaleste/sat_multiple_choice_math_may_23`.
 
+Our prompt is taken from from appendix G of Lewkowycz et al. (2022). 
 """
 
 from lm_eval.base import Task, rf
@@ -142,15 +137,6 @@ def end_seq(self):
  return ["\n\n", "Problem:"]
 
  def process_results(self, doc, results, params={}):
- """Take a single document and the LM results and evaluates, returning a
- dict where keys are the names of submetrics and values are the values of
- the metric for that one document
-
- :param doc:
- The document as returned from training_docs, validation_docs, or test_docs.
- :param results:
- The results of the requests created in construct_requests.
- """
  candidates = results[0]
  assert isinstance(params, dict)
  if params == {}: