Contrastive & dual encoders

So wird trainiert

Batch contrastive objectives reward paired examples and treat other examples as negatives. Negative quality, duplicate semantics, and temperature materially affect the learned space.

So läuft die Inferenz

Encode each side independently and compare vectors with dot product or cosine similarity. Precompute the large candidate side; optionally rerank top results with a cross-encoder.

Stärken

• •Scales retrieval to large corpora
• •Aligns modalities without a joint decoder
• •Enables zero-shot classification through label text in some settings

Zielkonflikte

• •Independent encoding misses fine-grained cross-input interactions
• •False negatives can distort training
• •Global similarity may ignore spatial, temporal, or compositional details

Geeignet, wenn

• Fast retrieval or matching is the first stage
• One side can be indexed offline
• Recall is followed by task-appropriate reranking when needed

Vermeiden oder hinterfragen, wenn

• Every candidate needs deep token-to-token comparison
• The task depends on precise spatial relationships
• Raw similarity scores are assumed to be probabilities