Archive
新模型、開放權重、多模態、語音、影像與模型架構。
最新模型
H3 Super Acceleration first creates a low-resolution draft with four denoising steps, then uses LTX and Sol-Attn to complete high-resolution refinement in three steps. NVIDIA reports that 768p, 10-second video generation is 27.7× faster than the SGLang baseline, but the acceleration changes the sampling path and is not a pure kernel optimization at equivalent quality.
代理工具
Microsoft has introduced an Agent Lightning Skill for Claude Code, Codex, and GitHub Copilot that lets a coding agent modify another editable agent against a fixed benchmark. It provides a budget-constrained set of experimentation rules; it does not automatically perform reinforcement learning on the target agent.
推論系統
The new architecture adds a low-cost model call between retrieval and the primary model, retaining only verbatim passages relevant to the question and their source identifiers. AWS’s internal tests recorded a roughly 30% cost reduction, but the evaluation relies on a single enterprise corpus and an LLM judge.
AI 安全
The research team used eight open model families to generate roughly 200,000 examples of pressure-induced scenarios, then applied LoRA to train models to determine whether they themselves were lying. The detectors mainly learned surface-level patterns such as sycophancy and capability denial, performing only slightly better than prompting baselines on unseen lie types.
最新模型
ReWorld assigns short-term action control and long-term scene memory to different attention ranges, then preserves history in a fixed-capacity landmark bank. The authors demonstrate real-time streaming at 704×1280 and 64-second return trajectories, but only the paper and demo page are currently available; neither model weights nor code have been released.
GitHub Repo
Prime Agent supports long-running coding agents with a persistent IPython environment, daemon sessions, and a versionable Continual Harness. Its authors report that ARC-AGI-3 RHAE Best@1 rose from 30% to 95.5%, but the score reflects the complete model-and-harness system and should not be treated as an isolated comparison of underlying model capabilities.
代理工具
The new benchmark uses private scientific tests to check units, numerical invariants, file semantics, and cross-module data flow, rather than merely whether code compiles. Claude Code paired with Opus 5 achieved an overall pass@1 of 47.90%, while experiments also showed that adding domain knowledge can sometimes anchor agents in the wrong direction.
GitHub Repo
DFlash 2 improves block-diffusion drafting with a candidate-path selector and dynamic convolution, delivering 2.7× to 3.4× the throughput of autoregressive decoding for Qwen3.8-27B in single-request tests. The latest reproducible issues show that vLLM nightly instantiates the wrong decoder layer and temporarily allocates nearly 4.74 GiB of extra memory before sharing vocabulary weights, potentially preventing startup on 24 GB GPUs.
硬體
Meta has disclosed the system design of MTIA 300, which accelerates recommendation-model training with built-in NIC chiplets, near-memory reduction units, and compiled HCCL. The company says communication time across 40 accelerators was reduced to roughly one-quarter that of a reference GPU cluster, but provided no comparisons of cost, power consumption, or general-purpose LLM training.
代理工具
The `latest` tag for DeepSeek Harness’s main package points to 0.1.1-rc.2, but several underlying packages still mark early versions as the latest. Installing the plugin API directly may trigger peer dependency conflicts or silently create a mismatched dependency tree.
AI 研究
η-learning constrains generative models with extreme value statistics to synthesize high-resolution extreme scenarios from paired data that lack rare events. The method can support infrastructure stress testing, but it cannot predict when disasters will occur and still depends on trustworthy external statistics.
代理工具
The Rails Foundation has released lemans, the Ruby evaluation framework used for the Agents on Rails leaderboard, with support for Daytona and local Docker sandboxes. The framework isolates validators, preserves replayable evidence, and restores protected files before scoring.