task_lite_clean_en/124/metadata.json

{
  "absolute_id": 124,
  "persona": "Researcher",
  "task": "The files `ScreenShot_2026-02-07_132218_499.png` and `ScreenShot_2026-02-07_132225_683.png` on my Desktop are screenshots I took while studying a paper. Please find that paper in the paper folder, extract its abstract, and output it as `output.txt`.",
  "task_diff": "hard",
  "output_files": [
    "output.txt"
  ],
  "rubrics": [
    "Is the paper title \"CL-BENCH: A benchmark for context learning\" included in the output.txt file?",
    "Does the output.txt file contain the full abstract section title?",
    "Does the summary extracted in output.txt start with \"Current language models (LMs) excel at reasoning over prompts using pre-trained knowledge\"?",
    "Does the summary mention that CL-bench contains 500 complex contexts, 1,899 tasks, and 31,607 validation criteria?",
    "Does the summary mention the evaluation result that ten frontier language models solved only 17.2% of tasks on average?",
    "Does the abstract mention that the best-performing model, GPT-5.1, solved only 23.7%?",
    "Does the end of the summary mention CL-bench as a step to build context-learning-competent language models that advance their deployment in real-world scenarios?",
    "Is the paper entitled CL-Bench: A Benchmark for Context Learning correctly identified from the two screenshots?",
    "Was the correct paper file `CL-Bench.pdf` matched instead of another paper file?",
    "Does the extracted abstract completely cover all paragraphs of the original abstract without missing key information?",
    "Does the extracted abstract accurately state that context learning is a key capability naturally possessed by humans but largely overlooked by current research?",
    "Abstract Is it accurate to say that solving tasks in CL-bench requires the model to learn from the context of new knowledge, rule systems, and complex procedures not found in the training data?",
    "Did you successfully read two png screenshot files to identify the thesis information?",
    "Was the `CL-Bench.pdf` paper file successfully read and was the abstract location correctly found?",
    "Was the extracted summary successfully output in txt format?",
    "Does the abstract make a clear distinction between CL-bench far beyond long contextual tasks and traditional contextual learning tasks?",
    "Does the summary mention that all tasks and assessment criteria are designed by experienced domain experts?"
  ],
  "rubric_types": [
    "Basic Evaluation",
    "Basic Evaluation",
    "Basic Evaluation",
    "Outcome Evaluation",
    "Outcome Evaluation",
    "Outcome Evaluation",
    "Outcome Evaluation",
    "Process Evaluation",
    "Process Evaluation",
    "Outcome Evaluation",
    "Outcome Evaluation",
    "Outcome Evaluation",
    "Process Evaluation",
    "Process Evaluation",
    "Basic Evaluation",
    "Outcome Evaluation",
    "Outcome Evaluation"
  ],
  "file_dep_graph": [
    {
      "from": "ScreenShot_2026-02-07_132218_499.png",
      "to": "CL-Bench.pdf"
    },
    {
      "from": "ScreenShot_2026-02-07_132225_683.png",
      "to": "CL-Bench.pdf"
    }
  ],
  "data_manifest": [
    {
      "filename": "CL-Bench.pdf",
      "stored_relpath": "data/bc80a4360a7d761b_CL-Bench.pdf"
    },
    {
      "filename": "ScreenShot_2026-02-07_132218_499.png",
      "stored_relpath": "data/748bf0f4e01ae135_ScreenShot_2026-02-07_132218_499.png"
    },
    {
      "filename": "ScreenShot_2026-02-07_132225_683.png",
      "stored_relpath": "data/8aa7f1d7b4d208b7_ScreenShot_2026-02-07_132225_683.png"
    }
  ],
  "tested_capabilities": [
    "Workspace Exploration",
    "Lineage Tracing",
    "Semantic Content Relations Understanding",
    "Heterogeneous File Understanding"
  ]
}