Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
Clark Hash achieves 32x compression of sentence embeddings by applying a deterministic sparse signed Johnson-Lindenstrauss projection, clipping, and scalar quantization, without requiring training or corpus statistics. This method represents 384-dimensional sentence embeddings with only 48 bytes while maintaining high cosine similarity accuracy. Experiments on multilingual sentence similarity tasks (STS17 and STS22) demonstrate strong performance, achieving macro Pearson correlations of 0.910 and 0.946 with a multilingual MiniLM encoder.
Store your sentence embeddings in 48 bytes without training, and still get >0.9 Pearson correlation on STS benchmarks.
Clark Hash is a small method for storing neural embeddings in less space. It normalizes each database vector, applies a deterministic sparse signed Johnson-Lindenstrauss projection, clips the result, and stores a fixed-width scalar-quantized code. Queries stay in floating point and are scored against the stored sketches. In the default 384-dimensional sentence-embedding setting, Clark Hash stores a cosine-search vector in 48 bytes instead of 1536 bytes for dense f32 storage. This is 32x smaller. The method does not need a training pass, learned codebooks, rotations, or corpus statistics before new vectors can be stored. We describe the codec, the Rust implementation, and a multilingual sentence-similarity evaluation on 9,304 labeled pairs from 29 subsets. With a multilingual MiniLM encoder, the 48-byte sketches reached 0.910 and 0.946 macro Pearson correlation with dense cosine scores on STS17 and STS22. Clark Hash is not a new Johnson-Lindenstrauss theorem and it is not a replacement for approximate nearest-neighbor indexes. It is a simple stateless codec for compact embedding storage.