[good first issue] Cross-model evaluation harness

[federicodeponte]

Build a harness to compare output quality across LLM providers.

What to do:

• Create scripts/eval_cross_model.py
• Run the same topic across Gemini, OpenAI, and Claude
• Compare: citation verification rate, generation time, cost
• Output a markdown comparison table

Skills needed: Python, API client experience

Related: EVALUATION.md

[PriyanshRaj30]

Hi! I'd like to work on this issue.

I'm currently reviewing the evaluation framework and existing provider integrations. I'll share an implementation plan shortly and start working on a draft solution.

[PriyanshRaj30]

I've implemented an initial version of eval_cross_model.py that evaluates providers through the full OpenDraft pipeline (generate_draft, CitationValidator, and TokenTracker)

[PriyanshRaj30]

Hi! I've opened a draft PR for this issue and implemented an initial version of eval_cross_model.py. I'd appreciate feedback on the evaluation approach, particularly around provider support and citation verification metrics.