Mathis Embit

Mathis Embit

AI Research Engineer

DeepSeek-R1: a Pure Reinforcement Learning Approach to Reasoning

DeepSeek-R1: a Pure Reinforcement Learning Approach to Reasoning

On January 20, 2025, DeepSeek unveiled DeepSeek-R1—a state-of-the-art, open-source reasoning model family. What distinguishes this release …

Read More
SPLADE: Combining Sparse and Semantic Approaches

SPLADE: Combining Sparse and Semantic Approaches

In Information Retrieval (IR), current research focuses on leveraging neural networks to find documents relevant to a query. Earlier …

Read More
Sparse Embeddings Demystified: From One-Hot Encoding to BM25

Sparse Embeddings Demystified: From One-Hot Encoding to BM25

This blog article aims to clarify some Information Retrieval (IR) methods related to sparse embeddings (or sparse vectors). We will study …

Read More
Reranking in RAG: Enhancing Accuracy with Cross-Encoders

Reranking in RAG: Enhancing Accuracy with Cross-Encoders

Reranking has became a must-have in a Retrieval-Augmented Generation (RAG) stack. This powerful technique enables filtering results from a …

Read More
Anthropic launches the Claude Enterprise plan with 500K context window

Anthropic launches the Claude Enterprise plan with 500K context window

On September 4, 2024, Anthropic has launched the Claude Enterprise plan. According to Anthropic’s announcement, the aim of the Claude …

Read More