embedding

Xetrieval is a mechanistic framework that explains dense retrieval by enhancing sentence embeddings with reasoning information and decomposing them into interpretable sparse features, providing feature-level explanations for retrieval decisions without expensive autoregressive generation.

Unveil introduces a unified visual-textual embedding framework for multi-modal document retrieval, using knowledge distillation to transfer semantic understanding from a visual-textual model to a purely visual model, achieving robust and efficient retrieval.

Garry Tan released new gbrain-evals benchmarks showing ZeroEntropy.dev achieves state-of-the-art performance in reranking and embedding cost, speed, and retrieval success, beating MemPalace and Vector RAG.

This paper presents an embedding-based federated learning pipeline for predicting iron deficiency from routine blood count data, deployed across two clinical sites with non-IID distributions. It demonstrates that personalized aggregation (FedMAP) outperforms standard FedAvg and local-only training, achieving higher ROC-AUC at both sites.

A tweet suggests that scaling the embedding learning rate by model width can replace the need for µP (micro-parameterization), referencing Muon optimizer for hidden layers and Adam for the rest.

LiquidAI releases LFM2.5-ColBERT-350M, a late-interaction multilingual retrieval model, along with a dense bi-encoder variant, both built on LFM2.5-350M-Base, supporting 11 languages and designed as drop-in replacements for RAG pipelines.

Introduces ReElicit, a Bayesian optimization framework that uses LLMs to elicit and adapt feature spaces for optimizing system prompts under aggregate scalar feedback, achieving strong performance across ten benchmark tasks.

This paper introduces m3BERT, a multilingual bidirectional encoder with a novel pretraining strategy that jointly optimizes representations across transformer layers and multiple embedding dimensions, enabling a single model to be adapted to varied resource constraints. It significantly outperforms state-of-the-art models on the Bing-Click industrial retrieval dataset.

EmbGen is a synthetic data generation pipeline that reassembles corpora into entity-description pairs using embedding similarity to generate diverse QA pairs for fine-tuning small language models on specialized domains, showing significant improvements in factual accuracy.

A technique to make embedding models aware of number ordering by overriding tokenizer and MLM fine-tuning, achieving 59% accuracy on number sorting benchmarks.

Perplexity AI releases pplx-embed-v1-late-0.6b, a small ColBERT late-interaction embedding model for retrieval, fine-tuned from their existing embedding model and optimized for MaxSim scoring, now open-source on HuggingFace.

The user converted Nvidia's Llama-Embed-Nemotron-8B model to MLX format with fp16, 8-bit, 4-bit, and 2-bit quantizations, enabling in-process embedding loading on Apple Silicon via mlx-embeddings.

This paper proposes Embedding Temporal Logic (ETL), a temporal logic that monitors perception-based autonomous systems directly in learned embedding spaces, enabling specification of high-level perceptual concepts and achieving strong empirical agreement with ground-truth semantics.

Jina has released Jina-embeddings-v5-omni, the first unified multimodal embedding model supporting text, images, audio, and video. The model is available in Small and Nano versions, is backward compatible with existing indexes, and boasts strong performance. It is now available on Hugging Face and via the Jina API.

Jina AI has released jina-embeddings-v5-omni, a universal embedding model supporting text, images, audio, and video with back-compatible indexing capabilities.

This article demonstrates how to embed Yggdrasil Network as a library in Go applications, allowing for user-space IPv6 mesh networking without a daemon.

Liquid AI releases LFM2.5-Embedding-350M, a dense bi-encoder for multilingual retrieval supporting 11 languages, as a drop-in replacement for RAG pipelines.

React Email 6.0 is an open-source email editor that can be embedded directly into your own application.

Google DeepMind introduces AlphaEarth Foundations, an AI model that integrates petabytes of Earth observation data into unified embeddings to map and monitor the planet at 10x10 meter resolution. The model's compact representations enable efficient planetary-scale analysis for applications in food security, deforestation tracking, and environmental monitoring.

Jina CLIP v2 is an improved multimodal embedding model supporting 89 languages, high-resolution images, and flexible embedding dimensions, with 3% better performance than v1 and state-of-the-art results on multilingual benchmarks.