Auto Added by WPeMatico
Liquid AI just shipped LFM2.5-8B-A1B. It is an on-device Mixture-of-Experts (MoE) model built for tool calling. The model holds 8.3B total parameters but activates only 1.5B per token. That sparsity is what lets it run on consumer hardware. The release follows LFM2-8B-A1B, which Liquid AI team published earlier. LFM2.5 is a new family of hybrid […]
Perplexity AI’s research team reimplemented their Unigram tokenizer from scratch in Rust and open-sourced the code in pplx-garden, their inference technology repository. At production input lengths, the new encoder cuts p50 latency by roughly 5x versus the Hugging Face tokenizers crate, ~2x versus SentencePiece (C++), and ~1.5x versus IREE’s tokenizer (C), with zero steady-state heap
Researchers from Sakana AI and the University of Tokyo propose DiffusionBlocks. It trains transformer-based networks one block at a time. Training memory is reduced by a factor of B, where B is the number of blocks. Performance is maintained across diverse architectures. The Memory Problem in Neural Network Training End-to-end backpropagation requires storing intermediate activations
Speculative decoding is a technique for speeding up large language model inference. A small, fast draft model proposes several tokens. The large target model verifies them in parallel. If accepted, inference is faster. If rejected, the system falls back gracefully. EAGLE Team, vLLM Team, and TorchSpec Team has launched the EAGLE series including EAGLE 1,
Long-context inference makes the KV cache one of the main costs of serving LLMs. During autoregressive decoding, the cache grows with context length, batch size, and model depth. At high batch sizes and long contexts with 100K tokens across dozens of concurrent requests the KV cache consumes a large fraction of GPU memory. Compressing it
The Model Context Protocol has moved from Anthropic’s internal experiment to a de facto industry standard at a speed few integration protocols have matched. Since its launch in November 2024, MCP has grown explosively: OpenAI adopted it in March 2025, Microsoft announced support in Copilot Studio in March 2025, and by late 2025 combined Python
Best Authentication Platforms for AI Agents and MCP Servers in 2026 Read More »
Linear attention replaces the unbounded KV cache of softmax attention with a fixed-size recurrent state. This cuts sequence mixing to linear time and decoding to constant memory. The hard part is not what to forget. It is how to edit a compressed memory without scrambling existing associations. NVIDIA has released Gated DeltaNet-2, a linear attention
Tencent has released TencentDB Agent Memory, an open-source memory system for AI agents. The project ships under the MIT license. It targets a problem familiar to anyone shipping long-horizon agents: context bloat and recall failure. It is symbolic short-term memory along with layered long-term memory. It integrates with OpenClaw as a plugin and with the
Instruction-tuned language models refuse harmful requests. But which part of the model is actually responsible — and how does that mechanism get installed during training? A new research from Nous Research team takes a neuron-level look at this question. The Nous research team developed contrastive neuron attribution (CNA), a method that identifies the specific MLP
Attackers increasingly target the packages, editor extensions, and AI tool configs on developer machines and not just production systems. Perplexity has open-sourced an internal tool it uses to address this problem. Perplexity released Bumblebee on GitHub. The tool is a read-only inventory collector for macOS and Linux developer endpoints. It is written entirely in Go
