[BUG] eval_results_table drops Series predictions with a non-default input index

[!WARNING] Before submitting a PR, please make sure that:

• A maintainer has triaged this issue and applied the ready label
• This issue has no assignee
• No duplicate PR exists

PRs not meeting these requirements may be automatically closed.

Issues Policy acknowledgement

• I have read and agree to submit bug reports in accordance with the issues policy

Where did you encounter this bug?

Local machine

MLflow version

• Client: 3.16.1.dev0, current master at f6252a80548ac13962788ac0cbd73799e8d51366
• Tracking server: local SQLite tracking URI

System information

• OS: Windows 11 (10.0.26200)
• Python: 3.12.5

Describe the problem

mlflow.models.evaluate can compute row-level metric values correctly but log None predictions in eval_results_table when evaluation data has a non-default pandas index and a callable model returns a Series with the normal RangeIndex.

The table assignment aligns the Series by label, while the metric path consumes predictions by position. Partially overlapping or reversed labels can therefore attach predictions to the wrong rows instead of making them null.

Expected: prediction values are attached to input rows by position, consistent with metric computation.

Actual: the logged table can disagree with the metrics.

Tracking information

Not applicable beyond the self-contained local SQLite setup below.

Code to reproduce issue

import tempfile
from pathlib import Path

import pandas as pd

import mlflow
from mlflow.metrics import MetricValue, make_metric


def predict(frame):
    return pd.Series([101.0, 202.0])


def row_values(predictions, targets):
    return MetricValue(scores=predictions.tolist())


work_dir = Path(tempfile.mkdtemp(prefix="mlflow-index-repro-"))
mlflow.set_tracking_uri(f"sqlite:///{work_dir / 'mlflow.db'}")
data = pd.DataFrame(
    {"feature": [1, 2], "target": [101.0, 202.0]},
    index=[10, 20],
)

with mlflow.start_run():
    result = mlflow.models.evaluate(
        predict,
        data,
        targets="target",
        model_type="regressor",
        evaluators="default",
        evaluator_config={"log_model_explainability": False},
        extra_metrics=[
            make_metric(
                eval_fn=row_values,
                greater_is_better=True,
                name="row_values",
            )
        ],
    )

table = pd.DataFrame(**result.artifacts["eval_results_table"].content)
print(table[["feature", "outputs"]].to_string(index=False))
assert table["outputs"].tolist() == [101.0, 202.0]

Stack trace

 feature outputs
       1    None
       2    None
Traceback (most recent call last):
  ...
AssertionError

Other info / logs

A local two-file patch and native regression coverage are ready if maintainers agree with this direction.

Willingness to contribute

Yes. I would be willing to contribute a fix with guidance from the MLflow community.

What component(s) does this bug affect?

• area/tracking
• area/model-registry
• area/scoring
• area/evaluation
• area/prompt
• area/tracing
• area/gateway
• area/projects
• area/uiux
• area/docs