Drill · lap 2 · 5/10 · RET-102 · A rank fusion that ranks the wrong way

[akash-coded]

5 of 10 · RET-102 · A rank fusion that ranks the wrong way

retrieval · medium · fix · ~12 min · no AWS account

Hybrid retrieval runs a lexical search and a dense search, then fuses the two rankings. Reciprocal rank fusion merges by rank, so you never have to reconcile two incompatible score scales. This implementation is subtly wrong, and the results look plausible enough to ship.

def rrf(rankings, k=60):
    """rankings: a list of ranked lists of doc ids, best first. Returns fused doc ids, best first."""
    scores = {}
    for ranking in rankings:
        for rank, doc_id in enumerate(ranking, start=1):
            scores[doc_id] = scores.get(doc_id, 0) + rank
    return sorted(scores, key=scores.get, reverse=True)

Find it

Run it in your head on [["a","b","c"], ["b","a","d"]]. Which document comes out on top — and should it?

The formula each retriever contributes for a document at rank r is 1 / (k + r). Fix the accumulation so a document ranked well by both retrievers rises, and one ranked well by only one is beaten by it.

/drill RET-102

```python
def rrf(rankings, k=60):
    ...
```

What this proves

That you understand why RRF uses a reciprocal — small ranks must produce large contributions — and that "the output is a sensible-looking list" is not evidence a ranker is correct.

---

Submit in this thread

Copy the block from the brief, edit it, and post it as a comment here. The bot replies within a couple of minutes with every check, the skill you demonstrated, and what to do next.

_{Brief in the repo: labs/drills/retrieval/RET-102 · next in the sequence: MAS-102 · how grading works · scoreboard}