Archive
GitHub 熱門專案、開源工具、框架、runtime 與重要 release。
AI 科學研究
OpenAI says its unreleased Astra model solved or advanced ten problems in mathematics and theoretical computer science and has released a 249-page manuscript. Each result includes a buildable Lean 4 formal certificate, but the mathematical significance and the manuscript’s arguments still await review by domain experts.
推論系統
Cloudflare uses an FP8 KV cache to expand Kimi K2.6’s concurrency capacity and switches to INT4 weights for GLM 5.2 decoding. The system also adds page-generation tags to the shared KV cache to prevent multi-tenant requests from reading incorrect cached data.
最新模型
MiniMax has released the weights for its H3 video model, enabling developers to run image-to-video generation with synchronized audio locally. ComfyUI has added an official workflow, but consumer GPU tests largely rely on pruned and INT8 derivative files and should not be treated as full-quality reproductions of the original weights.
AI 安全
CAGE does not merely inspect the tool response currently visible to an agent. It verifies that every plausible correct binding permits the same action. The research proves that checking categorical fields and numeric errors separately cannot guarantee safety when both change simultaneously.
代理工具
The compatibility period for `used_copilot_coding_agent` in GitHub Copilot usage reports expired on August 1. Integrations should migrate to `used_copilot_cloud_agent`. The two fields previously returned the same Boolean value, but the legacy field is no longer guaranteed after the deadline, potentially causing enterprise adoption reports to silently undercount usage.
GitHub Repo
NVIDIA, the Linux Foundation, Hugging Face, Microsoft, and other organizations have formed the Open Secure AI Alliance to address vulnerabilities in agents and AI software with open models and tools. Its first concrete contributions include NVIDIA’s NOOA framework, but the alliance has yet to announce common specifications, governance processes, or a delivery timeline.
AI 安全
New research suggests that the JSON-style tool specifications used by agents are not merely neutral interfaces and may suppress models’ internal refusal representations. By separately evaluating safety before execution using flattened text specifications, SafeKeep increased the average refusal rate for harmful requests from 23.8% to 70.6%.
代理工具
A Microsoft Research team jointly updates synthetic applications, tasks, and graders based on agent failure trajectories instead of merely adding more static webpages. The 9B model’s average success rate across 14 evaluations rose from 36.5% to 67.1%, although only four environments have been publicly released as benchmarks.
代理工具
Alibaba’s Qwen team has released a technical report on a GUI agent spanning mobile, desktop, web, and DeepSearch environments. Its unified action space combines on-screen interactions, CLI commands, and batched actions. The model leads existing systems on several self-reported benchmarks, but its weights, training data, and complete execution framework have not yet been released.
代理工具
Using 1,019 human-annotated trajectories spanning the web, Windows, Ubuntu, and mobile platforms, OSReward tests whether vision-language models can correctly determine whether computer agents have completed their tasks. The team also released 100,000 reasoning-based judgment examples and trained self-hostable 9B and 35B OS-Shepherd reward models.
最新模型
NVIDIA has released the Cosmos 3 Edge weights and inference pipeline, combining autoregressive reasoning and diffusion-based multimodal generation in a 4B Mixture-of-Transformers model. The model accepts text, images, and action trajectories, and a robot policy variant fine-tuned on DROID data is also available, though physical consistency and edge-device latency still require independent verification.
推論系統
WIDE brings dynamic pruning down to token-level attention-head groups and FFN channel groups, supporting both prefill and decode. At 50% sparsity, the study reports up to a 4.95× decoding speedup for core layers, but the end-to-end gain falls to 1.55×.