Gated DeltaNet-2
PulseAugur coverage of Gated DeltaNet-2 — every cluster mentioning Gated DeltaNet-2 across labs, papers, and developer communities, ranked by signal.
1 day(s) with sentiment data
-
New E-Gaborformer speech model offers smaller memory, lower error rates
Researchers have developed E-Gaborformer, a novel speech recognition model designed for real-time voice agents. This strictly causal encoder utilizes Gabor-dictionary convolutions and a delta-rule memory, resulting in a…
-
MemT-1.0 cuts LLM KV cache storage by 66x with memory-native design
A new paper, "MemT-1.0: Towards Memory-Native Evolution," introduces a novel approach to handling long-term memory in large language models. This method, called MemT-1.0, aims to reduce the significant memory and storag…
-
MemT-1.0 architecture integrates memory natively into foundation models
Researchers have introduced MemT-1.0, a novel architecture designed to make memory an intrinsic part of foundation models, enabling continuous information accumulation and utilization. This memory-native approach, built…
-
New QED method enhances long-range recall in linear attention models
A new research paper introduces Query-derived Erase Direction (QED), a method to improve long-range recall in linear attention models. QED adds a second erase direction derived from the query, orthogonal to the key, whi…
-
New research explores unified routing for adaptive LLM efficiency · 2 sources tracked
Two new research papers explore methods to optimize the efficiency of large language models by dynamically adjusting computational resources based on token complexity. The first paper, "Linear Attention Architectures," …
-
AI infrastructure firms Exa, Modal, Turbopuffer achieve major funding milestones
Several AI infrastructure companies have achieved significant funding milestones, with Turbopuffer reaching $100M ARR and profitability, Exa securing $250M in a Series C round valuing it at $2.2B, and Modal raising $355…
-
NVIDIA unveils Gated DeltaNet-2 for improved linear attention
NVIDIA has introduced Gated DeltaNet-2, a new linear attention layer designed to improve memory editing in recurrent neural networks. This model separates the processes of erasing old information and writing new informa…