calib·HumanEval/21
This is one (task, candidate response) pair flowing through the full PANOPTES pipeline. Each section below is a stage of the analysis: the task itself, the candidate solution being evaluated, every judge's score + rationale, the sampling-pass dispersion that captures within-judge noise, and the uncertainty-quantification metrics computed on top.
The function signature + docstring presented to both the model under test and to every judge. The entry-point name was obfuscated from rescale_to_unit to fn_9677d6b644 so the judges can't pattern-match a memorized HumanEval solution.
from typing import List
def fn_9677d6b644(numbers: List[float]) -> List[float]:
""" Given list of numbers (of at least two elements), apply a linear transform to that list,
such that the smallest number will become 0 and the largest will become 1
>>> fn_9677d6b644([1.0, 2.0, 3.0, 4.0, 5.0])
[0.0, 0.25, 0.5, 0.75, 1.0]
"""
The full solution every judge is grading. For these runs the candidate is the reference solution prepended with the task signature so it parses as a complete program.
from typing import List
def fn_9677d6b644(numbers: List[float]) -> List[float]:
""" Given list of numbers (of at least two elements), apply a linear transform to that list,
such that the smallest number will become 0 and the largest will become 1
>>> fn_9677d6b644([1.0, 2.0, 3.0, 4.0, 5.0])
[0.0, 0.25, 0.5, 0.75, 1.0]
"""
min_number = min(numbers)
max_number = max(numbers)
return [(x - min_number) / (max_number - min_number) for x in numbers]
One call per judge at temperature=0. This is the "best single guess" each judge has. Disagreement here is structural. The judges literally see this candidate differently.
n draws per judge at temperature=1. Dots are individual draws; vertical bar is the mean; shaded band is ±1σ. Wide band = the judge is uncertain even with itself; narrow band = it's consistent.
Three statistical methods computed on top of the raw scores above. Each one tells you something different about how much to trust the headline number.
How to read these: self-consistency = does the judge agree with itself when resampled? Narrower CI means more consistent. Semantic entropy = do the judge's rationales cluster into one meaning, or several? Higher entropy means the judge is internally conflicted about why. Decomposition = how much of the total variance comes from within-judge sampling noise (aleatoric, hard to fix) vs between-judge disagreement (epistemic, fixable by calling more judges).
The natural-language explanation each judge gave alongside its score. These are produced via tool-use structured output, so the score and the rationale are guaranteed to come from the same forward pass.
The solution correctly implements min-max normalization: it finds the min and max, then maps each element to (x - min) / (max - min). This produces the correct output for the given example [0.0, 0.25, 0.5, 0.75, 1.0]. The logic is sound for all typical cases. The only potential issue is a ZeroDivisionError when all elements are equal (max == min), but the task specifies "at least two elements" without guaranteeing they are distinct — however, this edge case is not explicitly addressed. The task says the list has at least two elements but doesn't guarantee distinct values, so this is a minor concern. The solution is clean, idiomatic, and correct for all realistic inputs described in the task.
The candidate solution correctly implements a linear transformation on the input list of numbers such that the smallest number becomes 0 and the largest becomes 1. The implementation uses the formula (x - min_number) / (max_number - min_number) for each element x in the list, which is the correct approach for normalizing the values between 0 and 1. The solution is idiomatic, clear, and handles edge cases such as lists with two elements, where the transformation is still valid. The provided example in the docstring is correctly handled by the implementation.