token embeddings
PulseAugur coverage of token embeddings — every cluster mentioning token embeddings across labs, papers, and developer communities, ranked by signal.
1 day(s) with sentiment data
-
New method uses adaptive prompts to boost time-series foundation models
Researchers have developed PaCTS, a novel method to enhance time-series foundation models (TSFMs) by using learned token embeddings as compact context surrogates. This approach generates instance-adaptive latent prompts…
-
New research reveals geometric biases in token embeddings and their impact on language model training
Three new arXiv papers explore the geometric and statistical properties of token embeddings in language models. The first paper identifies a "hub of short rows" near the origin of token embedding tables that inflates in…
-
New DEW technique offers robust text watermarking for LLMs
Researchers have developed a new text watermarking technique called Dual-Embedding Watermarking (DEW) designed for large language models (LLMs). This method uses both token-level and contextual embeddings, combined with…
-
New paper suggests LLMs learn causality via difference-making logic
A new paper proposes that large language models (LLMs) learn causal structure through a process called variational induction, which relies on identifying difference-makers within text data. The research argues that LLMs…
-
Spectral Retrieval enhances LLM agent localized search accuracy
Researchers have introduced Spectral Retrieval, a novel plug-in re-ranking stage for large language model (LLM) multi-agent systems. This method utilizes multi-scale sinc convolution over token embeddings to improve loc…