Archive
推論 runtime、量化、 serving、KV cache、成本與效能。
最新模型
DualG-MRAG uses a Macro Graph to find cross-document reasoning paths, then a Micro Graph to verify local image and text evidence, preventing fine-grained nodes from rapidly inflating the knowledge graph. In MMQA, the study raises R@5 from 42.1% for the strongest graph-based baseline to 61.9%, but the code and index construction costs have not yet been disclosed.
AI 安全
MAI-Cyber-1-Flash is a sparse MoE model with 137B total parameters, but activates only about 5B per token. It handles roughly 90% of MDASH’s workload before passing difficult cases to GPT-5.4. Microsoft says the new configuration cuts costs by nearly half, but the model is currently available only in private preview on Azure AI Foundry, and the benchmark results have yet to be independently reproduced.
代理工具
MANTA builds an agent communication graph for each task, then detects collaboration anomalies from execution traces and makes constrained changes to roles, connections, and verification order. Tested with Gemma 4 31B, it achieved an average score of 74.0 across five benchmarks, although the number of samples and repeated runs remained limited.
推論系統
A new round of similarly priced local AI testing shows that the M4 Max, with 546GB/s of unified memory bandwidth, leads the GB10 and Strix Halo in per-token decoding across three quantized models. The GB10 still has the advantage in prompt prefill and overall energy efficiency, demonstrating that tokens/s alone can misrepresent actual response times.
最新模型
SVR jointly trains answer generation, correctness judgments, and confidence estimation, allowing the model to decide at inference time whether to retain an answer or continue refining it. Qwen3.5-2B achieved 56.3% average accuracy across seven mathematics benchmarks, although its stopping signal may still be affected by the model’s own calibration biases.
開發者平台／API
GitHub shut down the Models playground, model catalog, inference API, and bring-your-own-key functionality on July 30, including for existing paying users. Applications that depend on its endpoints must migrate to Microsoft Foundry, another model gateway, or an in-house provider abstraction layer.
最新模型
A new study generated 686,000 synthetic textbooks totaling 32 billion tokens and isolated the effect of document structure through controlled experiments using identical content. Organizing related chapters into complete books outperformed splitting them into separate passages or concatenating them randomly, but the gains still come from the research team’s custom training pipeline, and neither the data nor the code has been released.
推論系統
SemPIC uses a separate Writer model to generate position-independent document KVs offline, while the Reader retains the original model and standard cache interface. Across three models and four tasks, average micro-F1 rises from 0.53 for existing methods to 0.60, but end-to-end latency, cache size, and code have not yet been disclosed.
推論系統
For models that generate future frames and robot actions simultaneously, QuantWAMs turns quantization calibration into a three-level decision process covering structure, joint objectives, and closed-loop trajectories. The simulated success rates of two models were within 0.2 to 0.7 percentage points of FP16, but the speed measurements cover only specific compute modules, while each real-robot comparison involved just ten trials per task.
代理工具
An OSWorld experiment shows that giving local GUI agents a small amount of screen history can substantially reduce repeated actions, but further extending the context or step limit mainly changes the types of failure. A two-stage planning architecture also trails a single agent because of formatting and coordination overhead, while generating more plans in parallel recovers only part of the gap at a high token cost.
最新模型
Moonshot AI built a taxonomy of perception failures from the earliest points of failure across 42 existing benchmarks, then used 3,000 short-answer questions to isolate ten basic visual capabilities. The results show that models with similar overall scores can have entirely different perceptual gaps, but because the test set was drawn from a larger internal question pool, contamination and evaluator bias remain concerns.
最新模型
Microsoft’s Mage-VL replaces uniform video frame sampling with I/P frames, motion vectors, and residual energy, encoding only regions with significant changes. Its 4B model reportedly reduces visual tokens by more than 75% while preserving performance on static tasks, although speed and quality comparisons still come primarily from official tests.