Google researchers published a January 2026 paper proving that a dual encoder's embedding dimension must grow linearly with the number of documents to express every possible ranking, while an autoregr…
A training-free, CPU-only lexical prompt-compression pipeline reduced prompt tokens by a mean of 40.3% (sigma = 9.2) while holding BERTScore-F1 at 0.876 against original-prompt output, according to an…
Google DeepMind researchers, with the University of Massachusetts Amherst and The University of Texas at Austin, published a paper titled "Autoregressive Ranking: Bridging the Gap Between Dual and Cro…
Researchers introduced AutoLexSteer, described as the first fully automated process for building steering vectors for large language models, alongside a new steering vector type called Hangman that op…
Qdrant published a September 08, 2026 article by Matin Mahmood and John Kupchanko explaining hyperbolic embeddings, which use negatively curved space to represent branching hierarchies more efficientl…
A developer created an open-source program to find the person with the funniest name by checking if first and last names are valid English words, using a dataset of nearly 500 million names from Faceb…
Researchers introduced GeoISF, a large-scale LiDAR-to-satellite image geo-localization pipeline that uses an instance semantic forest built from WordNet to bridge the modality gap between point clouds…
A new knowledge representation architecture called Synapedia defines concepts using events sourced from Wikipedia, addressing a structural flaw in existing systems like WordNet and Wikidata. The syste…
Researchers developed a methodology to transform Hindi WordNet into 1.25 million instruction-response pairs, fine-tuning a 12B-parameter language model to create a specialized conversational AI for lo…
Princeton graduate Nora Graves combined computer science and linguistics in her senior thesis, comparing large language models and statistical methods for adding accent marks to Yorùbá text. The stati…
An engineer built a documentation Q&A engine that returns `null` instead of hallucinating answers when a query has no match in the corpus. The Knowledge Base API uses BM25 retrieval with POS-aware lem…