Archive
AI 晶片、GPU、資料中心、邊緣裝置與硬體安全。
代理工具
NVIDIA has updated its Nemotron IMO 2026 resources, releasing SFT/RL weights, datasets, prompts, and complete resumable inference code. The system scored 30/42, but the result depends heavily on three 561B-parameter checkpoints and massive test-time compute, so it should not be treated as the capability of a model answering in a single pass.
最新模型
The NASA-IBM Lunar Foundation Model integrates more than 30 spatially aligned data layers from nine instruments across four missions and releases its weights under Apache 2.0. Official downstream experiments report reductions of up to 22% in ice-prediction error, but the results still come from the development team and have not yet been independently reproduced.
AI 安全
Datasette 1.0a39 and 0.65.4 patch multiple paths that could leak private data, involving case-variant table names, full-text indexes, relationship queries, and caching. The vulnerabilities were discovered with help from an external researcher and three frontier models, but the maintainers have deliberately withheld some tests for now, so operators should still treat upgrading as a priority.
AI 基礎設施／低精度訓練
NVIDIA’s low-precision training library adds Rubin SM107a support, two-dimensional MXFP8 weight scaling, and MoE communication paths. The release also disables a cuBLAS 13.7 algorithm that can silently corrupt data in grouped GEMM operations on B300 and Rubin.
最新模型
Agnes AI has relabeled the 33B weights on Hugging Face as Preview, explicitly stating that they are not the API model tested by Artificial Analysis. The two differ in context length, parameter disclosure, and benchmark results, so deployers can no longer use the leaderboard score of 36 to endorse the open checkpoint.
代理工具
autofinetune lets an agent modify Tunix training code according to a Markdown specification, run evaluations, and retain only Git commits that improve the target metric. Its initial experiments cover SFT for FunctionGemma and GRPO for Gemma 3 1B, but the workflow remains a small-scale, single-metric exploration.
最新模型
SenseNova-U1.5 uses a Mixture-of-Transformers (MoT) architecture that does not rely on a separate vision encoder or VAE, enabling a single model to handle understanding, reasoning, generation, and editing. The technical report added on September 11 details its training methods and evaluations. Existing weights and inference code are available, but the SFT, RL, and multi-expert distillation code has not yet been fully released.
代理工具
The new `claude plugin eval` command compares performance with a plugin enabled and disabled, producing JSON and HTML evaluation reports. Initial testing suggests that automatically generated criteria are better at verifying instruction following than proving that a plugin genuinely improves task quality.
推論系統
RunningHub has published a multi-GPU inference recipe for MiniMax H3 combining SGLang, step distillation, SageAttention2, Cache-DiT, and `torch.compile`. It reduces the reported latency for a five-second video from 348.8 seconds to 28.7 seconds, but the critical internal acceleration weights have not been released, and the public LoRA cannot directly reproduce that result.
最新模型
TinyDiT trains a 210M-parameter DiT from scratch on a single RTX PRO 6000 in 400,000 steps over 3.5 days, while releasing its code, weights, and stage-by-stage measurements. The results show that flow-matching loss is a poor proxy for image-generation progress, whereas registers, timestep shifting, and compiled training provide more actionable signals.
AI 基礎設施
Together AI ported its open-source CUDA kernel framework to NVIDIA Vera Rubin, reorganizing data reuse and the memory pipeline. Tests with square 16K matrices reached 22.24 PFLOPS, though the figure comes from a qualification sample and the still-maturing CUDA 13.4.
AI 安全
Anthropic’s new threat report shows that several types of attackers have upgraded Claude from a coding assistant into a schedulable attack-orchestration layer with persistent memory. Some workflows automatically modify, rebuild, and redeploy malware after detection, although humans still largely control target selection and the review of stolen data.