PANOPTES
panoptes-696da4d5

HE/2

humaneval · 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
3
sampling draws
15
1. the task

The function signature + docstring presented to both the model under test and to every judge.



def truncate_number(number: float) -> float:
    """ Given a positive floating point number, it can be decomposed into
    and integer part (largest integer smaller than given number) and decimals
    (leftover part always smaller than 1).

    Return the decimal part of the number.
    >>> truncate_number(3.5)
    0.5
    """
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 truncate_number(number: float) -> float:
    """ Given a positive floating point number, it can be decomposed into
    and integer part (largest integer smaller than given number) and decimals
    (leftover part always smaller than 1).

    Return the decimal part of the number.
    >>> truncate_number(3.5)
    0.5
    """
    return number % 1.0
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-haiku
1.000
$0.0026
claude-sonnet
1.000
$0.0068
gpt-4o-mini
1.000
$0.0002
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.

decomposition
aggregate (all judges)
Var = 0.00e+0
total predictive variance, 3 judges
aleatoric (irreducible)
0.00e+0
epistemic (reducible)
0.00e+0
self-consistency
claude-haiku
1.000[1.000, 1.000]
posterior mean · 90% Bayesian-bootstrap CI
variance: 0.00e+0
n samples: 5
self-consistency
gpt-4o-mini
1.000[1.000, 1.000]
posterior mean · 90% Bayesian-bootstrap CI
variance: 0.00e+0
n samples: 5
self-consistency
claude-sonnet
1.000[1.000, 1.000]
posterior mean · 90% Bayesian-bootstrap CI
variance: 0.00e+0
n samples: 5
semantic-entropy
claude-haiku
H = 0.000/ log 51.609
1 semantic cluster via bidirectional NLI
entropy as fraction of max
cluster sizes
c1=5
semantic-entropy
gpt-4o-mini
H = 0.000/ log 51.609
1 semantic cluster via bidirectional NLI
entropy as fraction of max
cluster sizes
c1=5
semantic-entropy
claude-sonnet
H = 0.000/ log 51.609
1 semantic cluster via bidirectional NLI
entropy as fraction of max
cluster sizes
c1=5

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-haiku
score 1.000 · 1786 tokens · 2047ms

The solution correctly implements the required functionality. The modulo operator (%) with 1.0 is the idiomatic Python approach to extract the decimal part of a positive floating-point number. It returns the remainder after dividing by 1, which is precisely the fractional component. The solution handles the stated example (3.5 → 0.5) correctly and will work correctly on edge cases like 0.0 (returns 0.0), 1.0 (returns 0.0), and other positive floats. The code is concise, clear, and efficient.

claude-sonnet
score 1.000 · 1729 tokens · 2929ms

The solution uses `number % 1.0`, which correctly returns the fractional/decimal part of a positive floating point number. It passes the canonical example (3.5 → 0.5) and handles edge cases like whole numbers (e.g., 3.0 → 0.0) and numbers close to an integer boundary correctly. The implementation is idiomatic and concise.

gpt-4o-mini
score 1.000 · 911 tokens · 2495ms

The candidate solution correctly implements the function to return the decimal part of a positive floating point number using the modulus operator. It passes the provided example and handles edge cases such as integers (e.g., truncate_number(3.0) returns 0.0) and very small numbers (e.g., truncate_number(0.1) returns 0.1). The code is clear and idiomatic.