Latest works
| Cover image | Cover image (dark) | Cover image (dark) | |||||
|---|---|---|---|---|---|---|---|
Tarka Embedding 30M V1 | Release | 7-1-2026 | We achieved 20× compression and recovered ~86% MTEB performance. The model supports elastic upscaling at inference, avoiding full-weight loading and reducing both memory footprint and compute cost. | Open - Source | tarka-embedding-30m-v1 | Grey Elegant Luxury Jewelry Boutique Business Card (6).png | ||
Reduce and Refine | Release | 1-12-2025 | This work explores model compression by progressively reducing a 28-layer model to a lean 6-layer model without major performance loss. Reduce and Refine demonstrates a practical path to faster, lighter, more efficient LLMs. | Open - Source | reduce-and-refine-6-equal-to-28 | reduce and refine.png | ||
Tarka Embedding V1 | Release | 9-11-2025 | The Tarka Embedding V1 series is a compact and efficient text embedding model family developed to explore the capabilities of knowledge distillation, coreset selection, and model compression techniques | Open - Source | tarka-embedding-v1 | Grey Elegant Luxury Jewelry Boutique Business Card (2).png | Grey Elegant Luxury Jewelry Boutique Business Card.png | Grey Elegant Luxury Jewelry Boutique Business Card (3).png |
.png?alt=media&token=6b1bb4cc-1b6d-4bd4-962c-fbbc0fa1f380){kind=link}
.png?alt=media&token=3fa9d02e-c804-464a-aa38-b66a0af086e3){kind=link}
.png?alt=media&token=9ecd3a10-a47b-47a8-b307-109d9267bd6a){kind=link}