Spaces:
Sleeping
Sleeping
File size: 1,627 Bytes
aab0192 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 | """
tasks/task_easy.py -- Easy task: 2 variables, low noise.
The agent must discover a single causal relationship between two abstract
variables in a low-noise setting with a generous budget.
Grader returns 0.0-1.0 based on:
- Hypothesis accuracy (60%)
- Efficiency bonus (20%)
- Calibration (20%)
"""
from __future__ import annotations
from typing import Any
TASK_EASY = {
"id": "easy",
"name": "Easy -- Single-Edge Discovery",
"description": (
"Discover the causal relationship between two abstract variables. "
"Low noise (sigma=0.05), generous budget (12 steps)."
),
"difficulty": "easy",
"reset_kwargs": {
"noise_level": "low",
"domain": "system_alpha",
"seed": 42,
},
}
def grade_easy(episode_result: dict[str, Any]) -> float:
"""
Grade an easy-task episode. Returns a score in [0.0, 1.0].
Args:
episode_result: Dict containing at minimum:
- accuracy_score (float): from the rubric
- efficiency_bonus (float): from the rubric
- calibration_score (float): from the rubric
- total_episode_reward (float): sum of all rubric components
Returns:
Normalized score between 0.0 and 1.0.
"""
accuracy = episode_result.get("accuracy_score", 0.0)
efficiency = episode_result.get("efficiency_bonus", 0.0)
calibration = episode_result.get("calibration_score", 0.0)
raw = (
0.60 * min(accuracy, 1.0)
+ 0.20 * min(efficiency / 0.15, 1.0)
+ 0.20 * min(calibration / 0.20, 1.0)
)
return round(max(0.0, min(1.0, raw)), 4)
|