fix

lm_eval/tasks/ai2d/ai2d.yaml

 dataset_path: lmms-lab/ai2d
+task: ai2d
 test_split: test
 output_type: generate_until
-doc_to_image: !function utils.doc_to_image
-doc_to_text: "<image>{{question | capitalize}}
-{{% for option in options %}}
-{{loop.index | chr(64)}}. {{option | capitalize}}
-{{% endfor %}}"
-doc_to_target: "{{options[answer]}}"
+doc_to_image:
+  - image
+doc_to_text: "<image>Look at the scientific diagram carefully and answer the following question: {{question | capitalize}}\n{% for option in options -%}{{['A', 'B', 'C', 'D', 'E', 'F'][loop.index0]}}. {{option | capitalize}}\n{% endfor -%}
+
+Think step by step and finally respond to the question with only the correct option number as \"FINAL ANSWER\"."
+gen_prefix: "Let's think step by step."
+doc_to_target: "{{ ['A', 'B', 'C', 'D', 'E', 'F'][answer|int] }}"
 generation_kwargs:
   until: []
   temperature: 0.0
   do_sample: false
   max_gen_toks: 512
+filter_list:
+  - name: "strict-match"
+    filter:
+      - function: "regex"
+        regex_pattern: "(?:[`\\*_]*(?:FINAL ANSWER|Final Answer)[`\\*_]*)[:\\s]*[`\\*_]*([A-D])[`\\*_]*"
+      - function: "take_first"
 metric_list:
   - metric: acc
     aggregation: mean