Archive
1534 technical news articles
推論系統
Instead of using fixed offloading, FreeToken uses measured bandwidth to decide whether to transfer cache-missed experts to the GPU or execute them directly on the CPU. The paper also demonstrates a 753B model served by a single workstation GPU, but the full weights still require system memory; this should not be interpreted as “fitting all 753B parameters on one GPU.”
AI 安全
Enterprise customers can now use Mythos 5 to scan code repositories and receive CWE classifications, severity levels, confidence ratings, and recommended fixes. The model remains behind a controlled service, and fixes require human review, but Anthropic has not disclosed detection accuracy.
代理工具
Using persistent memory, a supervisor agent, and a text-grid interface, AVO enabled Claude Opus 5 to complete every public level within 6,624 environment actions. The score of 100 covers only the public set and should not be interpreted as passing the semi-private or private competition tests.
AI 安全
The new release forces some Claude Code web tool requests through the session network proxy and also fixes issues involving proxied environments, MCP reconnections, and sensitive input handling. It adds a Python SDK migration command, but Anthropic has not disclosed the vulnerabilities’ severity or the scale of impact.
生成式視覺與 4D 重建
The system uses a fixed-length reference context and view regrouping during diffusion to reduce structural drift when generating multiview videos in batches. The weights and inference pipeline are now public, but the fully open-source 4D Gaussian Splatting reconstruction and low-memory mode are not yet complete.
代理工具
The MCP maintainers are concentrating the next phase of specification work on five areas, including server-initiated events, agent identity, and progressive tool discovery. These items remain on the roadmap, while current SDK support for Tasks and extended notifications still has gaps.
最新模型
Marin has begun training a mixture-of-experts (MoE) model with 535B total parameters and approximately 23B active parameters per token. The project plans to use 11 GB200 NVL72 systems to process roughly 18T to 18.75T tokens. Configurations, scaling experiments, and real-time telemetry are public, but model quality, the long-context strategy, and the final token budget have yet to be finalized.
最新模型
An Imperial College London team simplifies self-supervised audio pretraining with a causal Transformer, stop-gradient, and cosine loss. The method scales consistently across six classification benchmarks, but has yet to demonstrate generalization to ASR, generation tasks, or non-AudioSet corpora.
最新模型
A Beijing Academy of Artificial Intelligence research team has proposed a three-stage post-training method that enables models to answer questions about a fixed document set without RAG. IAR outperformed standard SFT across all metrics in seven of eight experiments, although the results still depend on synthetic question-answer pairs, model judges, and model-specific hyperparameter tuning.
最新模型
DeepSeek has added image understanding to V4-Flash, supporting text and images through three compatible interfaces while retaining the existing Flash pricing. However, its officially reported agent benchmark results have yet to be independently reproduced, and neither the model weights nor the vision architecture has been disclosed.
代理工具
The new release integrates centralized agent skill storage, search, and caching into VikingBot, while adding per-user memory retrieval policies and persistent background imports. The upgrade also removes the experimental relations API and CLI commands, requiring existing integrations to migrate first.
推論系統
The new release processes checkpoint staging and CUDA Graph capture concurrently during startup, delivering a 2.38× speedup over the standard default workflow in official H100 testing. It also rewrites parts of the tensor-parallel communication path and adds seven model families, but the Torch upgrade and cache-path consolidation introduce migration costs.