Add 31 PRBench task specifications (instructions, task.yaml, paper markdown; ground truth withheld)
a48f2b1 verified | # PRBench Task Configuration | |
| # Each paper has one task.yaml that contains ALL task-specific information. | |
| # Agent code is fully generic and reads everything from this file. | |
| task_id: task_fu_2001 | |
| # --- Paper Info --- | |
| paper: | |
| title: "Classical collisional trajectories as the source of strong-field double ionization of helium in the knee regime" | |
| author: "L. B. Fu, J. Liu, J. Chen, S. G. Chen" | |
| doi: "10.1103/PhysRevA.63.043416" | |
| year: 2001 | |
| paper_file: "fu2001.md" # relative to task dir | |
| # --- Files --- | |
| instruction_file: "instruction.md" # what the white agent receives (no ground truth) | |
| metadata_file: "metadata.md" # ground truth formulas + methodology (for grading) | |
| ground_truth_data_dir: "data" # directory with reference CSV files | |
| ground_truth_code_dir: "reproduction" | |
| # --- Expected Outputs (what the white agent must produce) --- | |
| expected_outputs: | |
| analysis: | |
| - "reproduction/ANALYSIS.md" | |
| code: | |
| - "reproduction/fig3_knee_structure.py" | |
| data: | |
| - "data/fig3_knee_structure.csv" | |
| # --- Docker Environment --- | |
| docker: | |
| image: "python:3.11-slim" | |
| memory_limit: "4g" | |
| timeout: 21600 # 6 hours max (Monte Carlo over 13 intensity points) | |
| pip_install: | |
| - numpy | |
| - scipy | |
| - matplotlib | |
| # --- Grading Rubric (for model-judge) --- | |
| grading: | |
| dimensions: | |
| - name: "methodology_understanding" | |
| weight: 0.20 | |
| description: > | |
| Does the agent's ANALYSIS.md correctly describe the quasistatic CTMC model, | |
| ADK tunneling rate, microcanonical distribution for the inner electron, | |
| Kustaanheimo-Stiefel regularization, and ionization classification? | |
| Compare against metadata.md. | |
| - name: "code_correctness" | |
| weight: 0.15 | |
| description: > | |
| Is the CTMC implementation correct? Check laser field envelope, ADK weight | |
| computation, KS regularization, adaptive RK integration, and SI/DI classification | |
| based on single-particle energies. | |
| - name: "data_accuracy" | |
| weight: 0.50 | |
| description: > | |
| How well does the generated CSV match the ground truth reference data? | |
| CTMC is stochastic so allow factor-of-3 relative tolerance on individual points. | |
| Key checks: He2+/He+ ratio plateau at ~0.002 for I > 1e15 W/cm2, | |
| monotonically increasing DI yield, correct qualitative knee structure. | |
| - name: "completeness" | |
| weight: 0.15 | |
| description: > | |
| Is the CSV file, the code file, and the ANALYSIS.md present? | |
| Is the data file non-trivial (13 rows, correct column names)? | |
| # --- Code-Only Evaluation Mode --- | |
| code_only: | |
| instruction_suffix: | | |
| ## IMPORTANT: Code-Only Mode | |
| In this task you are in **code-only mode**. You must: | |
| 1. Read and analyze the paper | |
| 2. Write `reproduction/ANALYSIS.md` with your methodology analysis | |
| 3. Implement ALL code files under `reproduction/` | |
| 4. **DO NOT** run the Monte Carlo simulation or generate CSV data files | |
| 5. Focus on writing correct, complete, and well-structured code | |
| Your deliverables are ONLY the analysis document and code files. | |
| Do NOT produce any CSV data files. | |
| expected_outputs: | |
| analysis: | |
| - "reproduction/ANALYSIS.md" | |
| code: | |
| - "reproduction/fig3_knee_structure.py" | |
| grading: | |
| dimensions: | |
| - name: "methodology_understanding" | |
| weight: 0.25 | |
| description: > | |
| Does ANALYSIS.md correctly describe the two-step CTMC model: | |
| ADK tunneling (Step 1) and classical trajectory propagation (Step 2)? | |
| Are the KS regularization and microcanonical sampling correctly described? | |
| - name: "code_correctness" | |
| weight: 0.50 | |
| description: > | |
| Read the code carefully. Check whether the implementation is | |
| physically and algorithmically correct: | |
| - Total potential: V = -Z/r1 - Z/r2 + 1/|r1-r2| + (r1+r2)·F(t) | |
| - Laser field: F(t) = a(t)*F*cos(omega*t) with flat-top + cos^2 ramp | |
| - Microcanonical sampling at E2 = -2 a.u. | |
| - ADK weight: w = w^(0) * w^(1) where w^(1) has Gaussian transverse factor | |
| - KS map: x = u1^2-u2^2-u3^2+u4^2, etc., with dt/ds = r | |
| - SI: E1>0 AND E2<0; DI: E1>0 AND E2>0 | |
| - Yield: weighted sum DI/all | |
| Compare against ground truth formulas in metadata. Flag any bugs. | |
| - name: "code_quality" | |
| weight: 0.10 | |
| description: > | |
| Is the code well-structured and readable? Is the CTMC pipeline | |
| clearly separated into laser, sampling, propagation, and classification modules? | |
| - name: "completeness" | |
| weight: 0.15 | |
| description: > | |
| Are the code file and the ANALYSIS.md present and non-trivial? | |