[FEATURE] Add `expected_*` into trace-based evaluators.

[poshinchen]

Problem Statement

Similar to Output-based evaluators( OutputEvaluator, TrajectoryEvaluator), the trace-based evaluators should be able to take expected_* parameters into considerations for evaluation.

Trace-based evaluators:

1. faithfulness_evaluator
2. goal_success_rate_evaluator.py
3. harmfulness_evaluator.py
4. helpfulness_evaluator.py
5. response_relevance_evaluator.py
6. tool_parameter_accuracy_evaluator.py
7. tool_selection_accuracy_evaluator.py

Notes

• As of now, expected_output and expected_trajectory are the "expected similar" output, not the "exact" outputs. The criteria is further tuned in the rubric. We need to consider what the expected behavior is for trace-based evaluators.

Proposed Solution

No response

Use Case

N/A

Alternatives Solutions

No response

Additional Context

No response