Rag metrics

[assaftibm]

• New RAG metrics (details below)
• A new tool for executing evaluation metrics on an input DataFrame, and setting results as new columns in the DataFrame
• Updates to the Perplexity metric

List of added RAG metrics:

• reference-less:

  • context-relevance:

    • metrics.rag.context_relevance

  • faithfulness:

    • metrics.rag.faithfulness

    • metrics.rag.k_precision

    • metrics.rag.bert_k_precision

  • answer-relevance:

    • metrics.rag.answer_relevance

• reference-based:

  • context-correctness:
    • metrics.rag.mrr
    • metrics.rag.map
  • answer-correctness:
    • metrics.rag.correctness
    • metrics.rag.recall
    • metrics.rag.bert_recall

[matanor]

LGTM, left just one small question

[codecov]

Codecov Report

Attention: 7 lines in your changes are missing coverage. Please review.

Comparison is base (ab14e92) 94.18% compared to head (df0985a) 94.13%.

Files	Patch %	Lines
src/unitxt/eval_utils.py	41.66%	7 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #508      +/-   ##
==========================================
- Coverage   94.18%   94.13%   -0.06%     
==========================================
  Files         193      194       +1     
  Lines        7759     7803      +44     
==========================================
+ Hits         7308     7345      +37     
- Misses        451      458       +7     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.