cosine similarity
technique judged 100% likely a named technique or architecture
See how it has spread across media tiers
Mentions over time
Table view
| Week | Articles | Sources |
|---|---|---|
| 2024-W05 | 1 | 1 |
| 2024-W11 | 1 | 1 |
| 2024-W26 | 1 | 1 |
| 2026-W09 | 3 | 1 |
| 2026-W11 | 6 | 1 |
| 2026-W13 | 1 | 1 |
| 2026-W18 | 1 | 1 |
| 2026-W24 | 2 | 1 |
| 2026-W31 | 1 | 1 |
| 2026-W37 | 1 | 1 |
| 2026-W38 | 1 | 1 |
| 2026-W40 | 4 | 2 |
| 2026-W41 | 1 | 1 |
In their words
score # Max similarity to already selected max_sim = max ( cosine_similarity ( candidate_vectors [ idx ], candidate_vectors [ sel ] ) for sel in selected ) # MMR score mmr_score = lambda_mult * relevance - ( 1 - lambda_mult ) * max_sim mmr_scores .
My RAG bot had exactly one safety rule: if the best chunk scores below 0.8 cosine similarity, answer "I don't know." That cosine similarity threshold ran for months without complaint.
Many teams still ship the 2023 version: fixed chunks, embeddings, top-five by cosine similarity, hope.
Finding the right passage when the words don't match Pure keyword or basic cosine-similarity search misses conceptual synonyms and also "drawdown mitigation" and "capital preservation" mean roughly the same thing to a person asking a question, but not to a system doing literal string or frequency matching.
Detail a dynamic strategy: if a query fails, your architecture decreases the vector search similarity threshold (e.g., moving cosine similarity from 0.85 to 0.70) or switches from a dense vector search to a keyword BM25 search.
It compares the incoming label’s vector with the project’s existing pattern vectors using cosine similarity and reuses the nearest pattern if its score reaches MKG_TASK_PATTERN_SIMILARITY_THRESHOLD, which defaults to 0.80.
It's typically 10–30x cheaper and 20–50x faster than LLM-based routing because it uses a single embedding call and cosine similarity instead of a chat completion.
The result dict includes: • drift_phase: stable / shifting / drifted; a jump signals a topic switch (distinct from the six conversational state phases — initialization through instability) • drift_score: 0.0–1.0 numeric distance from the accumulated context • top_similarity: cosine similarity to the rolling top-K of recent embeddings • short_circuit: true when a paraphrase of an earlier question i