Technical AI news, automatically curated and generated
NVIDIA Introduces Nemotron-TwoTower-30B: A Hybrid Diffusion-Based Language Model NVIDIA has unveiled the Nemotron-TwoTower-30B-A3B-Base-BF16, a novel architectural approach to language modeling that integrates a diffusio…
→ View original source
Baidu Introduces Unlimited OCR: A 3B MoE Model Optimizing Long-Document Parsing via R-SWA Baidu has open-sourced "Unlimited OCR," a 3-billion parameter Mixture-of-Experts (MoE) model designed to eliminate the linear memo…
→ View original source
Beyond Translation: How Hi Translate 6.0 Is Evolving into a Multilingual AI Agent TRANSSION has announced the launch of Hi Translate 6.0, marking a strategic transition from a traditional translation tool to a comprehens…
→ View original sourceNVIDIA-NeMo Introduces Megatron-Bridge: Seamless Interoperability Between Megatron and Hugging Face NVIDIA-NeMo has released Megatron-Bridge, a specialized training library designed to bridge the gap between Megatron-bas…
→ View original sourceEvaluating VRAM Scaling for Large-Scale Model Inference: Performance on RTX 6000 PRO Clusters A technical discussion regarding the feasibility and performance of deploying massive models—specifically GLM 5.2, Kimi 2.7, a…
→ View original sourceNSA Loses Access to Anthropic's Mythos Following Corporate Dispute The National Security Agency (NSA) has reportedly lost access to "Mythos," a specialized tool developed by AI firm Anthropic, following a dispute between…
→ View original source
OpenAI and Broadcom Unveil Custom Silicon for Large-Scale LLM Inference OpenAI has partnered with Broadcom to develop a specialized chip architecture optimized for the inference of Large Language Models (LLMs) at scale, …
→ View original sourceAnthropic Alleges Illicit Extraction of Claude AI Model Capabilities by Alibaba Anthropic has raised serious allegations against Alibaba, claiming that the Chinese tech giant illicitly extracted capabilities from the Cla…
→ View original source
Constraint Tax in Open-Weight LLMs: An Empirical Study of Tool Calling Suppression Under Structured Output Constraints Researchers have identified a phenomenon termed "Tool Suppression," where the simultaneous enforcemen…
→ View original source
V-Zero: Answer-Label-Free On-Policy Distillation with Contrastive Evidence Gating for Fine-Grained Visual Reasoning V-Zero is a novel framework that enables fine‑grained visual reasoning in multimodal large language mode…
→ View original source
Causal-rCM: Advancing Autoregressive Diffusion Distillation for Streaming Video and World Models Researchers introduce Causal-rCM, a unified distillation framework that integrates teacher-forcing and self-forcing mechani…
→ View original source
Evaluating the Impact of Thinking Tokens on Model Safety and Alignment New research examines whether the "thinking tokens" used by modern reasoning models actually enhance safety and alignment, challenging the common ass…
→ View original source