PANOPTES
panoptes-e86ef9e3

calib·HumanEval/23

humaneval-calibprobe · code

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.

mean score
1.000
inter-judge spread
0.000
judges polled
2
sampling draws
20
1. the task

The function signature + docstring presented to both the model under test and to every judge. The entry-point name was obfuscated from strlen to fn_4719e7e500 so the judges can't pattern-match a memorized HumanEval solution.



def fn_4719e7e500(string: str) -> int:
    """ Return length of given string
    >>> fn_4719e7e500('')
    0
    >>> fn_4719e7e500('abc')
    3
    """
2. the candidate response judged

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.



def fn_4719e7e500(string: str) -> int:
    """ Return length of given string
    >>> fn_4719e7e500('')
    0
    >>> fn_4719e7e500('abc')
    3
    """
    return len(string)
3a. point-pass scores (temperature 0)

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.

claude-sonnet
1.000
$0.0066
gpt-4o
1.000
$0.0027
3b. sampling-pass dispersion (temperature 1)

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.

4. uncertainty quantification

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.

semantic-entropy
claude-sonnet
H = 0.000/ log 102.303
1 semantic cluster via bidirectional NLI
entropy as fraction of max
cluster sizes
c1=10
semantic-entropy
gpt-4o
H = 0.000/ log 102.303
1 semantic cluster via bidirectional NLI
entropy as fraction of max
cluster sizes
c1=10

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).

5. judge rationales

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.

claude-sonnet
score 1.000 · 1667 tokens · 2653ms

The solution correctly uses Python's built-in `len()` function to return the length of the string. It handles both the empty string edge case (returns 0) and normal strings (returns correct length). The implementation is idiomatic, concise, and correct for all plausible inputs.

gpt-4o
score 1.000 · 852 tokens · 2437ms

The candidate solution correctly implements the function to return the length of a given string using Python's built-in `len()` function. It passes the provided examples, including an empty string and a non-empty string, which are typical edge cases for this task. The solution is idiomatic, clear, and handles all plausible edge cases effectively.