metadata: 2026-07-07 revised taxonomy (task_type format vocabulary multiple_choice/yes_no/short_answer/free_form; score_type incl. none; re-verified protocol evidence; repaired prompt provenance)

metadata.json

@@ -9,7 +9,7 @@
       "modalities": [
         "single_image_start"
       ],
-      "task_type": "vqa",
+      "task_type": "short_answer",
       "score_type": "llm_judge",
       "score_params": {
         "rubric": "scale",
@@ -50,8 +50,8 @@
       },
       "prompt_template_source": {
         "origin": "official",
-        "reference": "https://github.com/EvolvingLMMs-Lab/lmms-eval/blob/main/lmms_eval/tasks/benchlmm/utils.py (benchlmm_doc_to_text — VLMEvalKit canonical short-answer trailer)",
-        "notes": "Tier 4: lmms-eval BenchLMM canonical evaluation prompt."
+        "reference": "https://github.com/AIFEG/BenchLMM/blob/main/baseline/LLaVA/BenchGPT_LLaVA_model_vqa.py#L44-L49 (qs = line['text']; DEFAULT_IMAGE_TOKEN + '\\n' + qs); trailer '\\nAnswer the question using a single word or phrase.' is embedded in every official jsonl 'text' field, e.g. https://github.com/AIFEG/BenchLMM/blob/main/jsonl/Benchmark_style_cartoon.jsonl",
+        "notes": "Tier 1: BenchLMM authors' own inference script feeds <image>\\n + question, and the official jsonl question text carries the '\\nAnswer the question using a single word or phrase.' trailer; the mm-eval template reproduces this (trailer re-appended because the mm-eval question column is trailer-free)."
       }
     }
   }