allow multiple rules

VerifiableRewardsForScalableLogicalReasoning.py

@@ -91,7 +91,7 @@ Args:
     references (`list` of `dict`): Each reference should contain:
         - 'validation_program' (`str`): Background knowledge in Prolog syntax
         - 'evaluation_config' (`dict`, optional): Configuration of predicates to use for evaluation.
-        Define: positive_predicate, and negative_predicate, the positive one should match the head of the rule to evaluate.
+        Define: positive_predicate, and negative_predicate, the positive one should match the head of the rule to evaluate. 
 Returns:
     accuracy (`float`): The proportion of predictions that correctly solve all examples. Value is between 0 and 1.
     partial_score (`float`): Average proportion of correctly classified examples across all predictions. Value is between 0 and 1.
@@ -130,9 +130,10 @@ def _evaluate_with_prolog(prediction, validation_program, eval_config, timeout=5
     # Extract configuration
     positive_pred = eval_config.get("positive_predicate", "eastbound")
     negative_pred = eval_config.get("negative_predicate", "westbound")
+    allow_multiple_rules = eval_config.get("allow_multiple_rules", True)
 
     # extract predicate from rule_to_evaluate
-    rule_to_evaluate = extract_ilp_from_text_v2(prediction, positive_pred)
+    rule_to_evaluate = extract_ilp_from_text_v2(prediction, positive_pred, allow_multiple_rules)
     if positive_pred not in rule_to_evaluate:
         logger.warning(f"Rule '{rule_to_evaluate}' does not contain positive predicate '{positive_pred}'")
         return {
@@ -241,16 +242,16 @@ def extract_ilp_from_text(text):
     return p_code
 
 
-def extract_ilp_from_text_v2(text, target_predicate=None):
+def extract_ilp_from_text_v2(text, target_predicate=None, allow_multiple_rules=False):
     text = re.sub(r'%.*?(?=\n|$)', '', text) # remove comments
     # Pre-process: collapse code blocks to single lines
     text = re.sub(r'\n\s*', ' ', text)  # crude: flatten all to one line
     # Rule pattern, across newlines
     rule_pattern = re.compile(rf'({target_predicate}\([^()]*\)\s*:-.*?\.)')
     rules = list(rule_pattern.findall(text))
-    if len(rules) > 1:
-        logger.warning(f"Found multiple rules in text: {rules}. Using only the first one.")
-        rules = rules[:1]  # Use only the first match
+    if len(rules) > 1 and not allow_multiple_rules:
+        logger.warning(f"Found multiple rules in text, but allow_multiple_rules is set to False. Using only the first match.")
+        rules = rules[:1]
     # Remove rules that are also captured as facts
     p_code = ''
     for rule in rules: