minilm
tool judged 3% likely a named technique or architecture
See how it has spread across media tiers
Mentions over time
Table view
| Week | Articles | Sources |
|---|---|---|
| 2022-W28 | 1 | 1 |
| 2024-W22 | 1 | 1 |
| 2024-W33 | 1 | 1 |
| 2026-W09 | 1 | 1 |
| 2026-W11 | 1 | 1 |
| 2026-W31 | 1 | 1 |
| 2026-W40 | 1 | 1 |
| 2026-W41 | 1 | 1 |
In their words
get ( " rerank_score " , r [ " score " ]), rank = i ) for i , r in enumerate ( final ) ] # Cross-encoder reranker (best quality) from sentence_transformers import CrossEncoder reranker_model = CrossEncoder ( " cross-encoder/ms-marco-MiniLM-L-6-v2 " ) def rerank ( query : str , documents : List [ str ]) -> List [ float ]: pairs = [( query , doc ) for doc in documents ] scores = reranker_model .
from sentence_transformers import SentenceTransformer from pinecone import Index model = SentenceTransformer ( ' all-MiniLM-L6-v2 ' ) index = Index ( ' my-docs ' ) def retrieve ( query , top_k = 5 ): vec = model .
The default is all-mpnet-base-v2 (768-d), which has better recall than all-MiniLM-L6-v2 (384-d).
Strong on retrieval tasks specifically - all-MiniLM-L6-v2 — 384 dimensions.
Both accept a query and a list of candidate texts and return relevance scores: For self-hosted reranking (lower latency, no API cost), the cross-encoder/ms-marco-MiniLM-L-6-v2 model from sentence-transformers is a strong open-source option that fits on a single GPU: Latency Tradeoffs Section titled “Latency Tradeoffs” Reranking 20 candidates with Cohere Rerank v3 adds approximately 150–300ms to th
We use cross-encoder/ms-marco-MiniLM-L-6-v2, which has a good balance between accuracy and inference latency.
We use cross-encoder/ms-marco-MiniLM-L-6-v2, which has a good balance between accuracy and inference latency.
All we need for the playlist generator is to find songs that have some semantic similarity, and since I don’t really care about hitting a particular performance metric I arbitrarily chose sentence-transformers/msmarco-MiniLM-L-6-v3.