splade cocondenser selfdistil

Name: splade cocondenser selfdistil
Author: NAVER

by NAVER

Open WeightModel weights are publicly available — can be downloaded and self-hosted

Released May 20221K context≈ 384 words

A specialist in sparse retrieval — this model produces sparse vector representations where most values are zero, making it efficient for keyword-aware semantic search. It uses a co-condenser architecture with self-distillation training, which helps it learn better term weighting without requiring massive labeled datasets. Its 512-token context limit keeps it focused on shorter passages rather than long documents.

Capabilities

Capability scores are AI-generated based on model documentation, benchmarks, and technical specifications. Learn more

Tool Use

Moderate

Instruction Following

splade cocondenser selfdistil

by NAVER

Open WeightModel weights are publicly available — can be downloaded and self-hosted

Released May 20221K context≈ 384 words

A specialist in sparse retrieval — this model produces sparse vector representations where most values are zero, making it efficient for keyword-aware semantic search. It uses a co-condenser architecture with self-distillation training, which helps it learn better term weighting without requiring massive labeled datasets. Its 512-token context limit keeps it focused on shorter passages rather than long documents.

Capabilities

Capability scores are AI-generated based on model documentation, benchmarks, and technical specifications. Learn more

Tool Use

Moderate

Instruction Following

Glossary

ArchitectureThe underlying structural design of a neural network that defines how data flows through layers and components.Co-Condenser ArchitectureA neural network design that jointly trains two components together to produce better embeddings by learning from both query and document representations simultaneously.DistillationA technique that compresses a large, complex model into a smaller one by training it to mimic the larger model's behavior, resulting in faster inference with minimal loss of quality.RetrievalThe process of finding and returning relevant documents or information from a database based on a query.Semantic SearchA search method that finds results based on the meaning of text rather than just matching keywords, using embeddings to understand intent.Sparse RetrievalA search method that represents text as a high-dimensional vector with mostly zeros, focusing on keyword matching and exact term overlap.Sparse VectorA vector representation where most values are zero, allowing efficient storage and fast similarity comparisons by only tracking non-zero positions.Term WeightingThe process of assigning importance scores to individual words or terms in a document, so that more relevant words have higher values in the embedding.TokenA small unit of text (a word, subword, or punctuation mark) that a language model breaks input into for processing.

Capabilities

Capabilities

Use Case Fit

Glossary