最新模型

新模型、開放權重、多模態、語音、影像與模型架構。

IAR Splits Document Internalization Into Injection, Question-Answer Alignment, and Weight Merging, Preserving More General Capabilities on Average

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.

2026-08-23

最新模型

DeepSeek V4-Flash-Vision-Exp Adds Image Input and a 1M-Token Context Window, but Remains an Experimental API

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.

2026-08-23

推論系統

SGLang 0.5.18 Overlaps Weight Loading with CUDA Graph Capture, Cutting Qwen3-32B Cold Start to 35.6 Seconds

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.

2026-08-23

代理工具

ASR Benchmark Fitting Quantified with Three Probe Types; Some Models Reproduce Erroneous Reference Transcripts 18% to 30% of the Time

Hume AI and Hugging Face found that when audio conflicts with the answers in public test sets, some high-scoring speech models still output content from the reference transcripts. The Open ASR Leaderboard has added a Benchmark Fitting tab, but these behaviors do not prove that the models directly encountered the test data.

2026-08-23

硬體

Needle 2 Compresses a 45M-Parameter Tool Model into 14MB and Constrains On-Device Calls with Schema Grammars

Cactus Compute recently improved Needle 2’s Python API, cross-platform execution, and Apache 2.0 licensing, while continuing to update its model card. It performs tool selection and structured extraction with 28MB of peak memory, although its speed and accuracy claims still come primarily from official tests.

2026-08-23

推論系統

Yutori Reveals Browser Agent Inference Architecture: 100K-Token Inputs, Short Tool Outputs, and Prefix Caching to Reduce Per-Step Costs

Yutori’s Navigator calls a vision-language model dozens of times during each browser task, with historical context growing to roughly 100K tokens. Together AI handles this workload using vLLM prefix caching and separate latency targets, although the claimed 2× speedup and 4–5× lower cost remain vendor-reported comparison results.

2026-08-23

代理工具

Slack Code Brings Agent Workflows into Shared Channels, but the Custom Agent API Is Not Yet Widely Available

Slack Code lets Claude, Devin, Copilot, ChatGPT, and Vercel agents present plans, code diffs, and live previews in project-specific channels. It improves the interface for multi-person agent oversight, but execution sandboxes, credentials, and repository permissions remain the responsibility of each agent provider.

2026-08-23

推論系統

LFM2.5 Adds DSpark Draft Models, Boosting H100 Decoding Throughput by Up to 3.18×

Liquid AI has released DSpark draft checkpoints with about 300 million parameters for three LFM2.5 models and integrated them with llama.cpp and SGLang. Official tests show that acceleration varies significantly by model, data distribution, and hardware, so peak results should not be treated as representative of gains across all workloads.

2026-08-22

最新模型

RuleMaze Breaks Visual Planning Into Tool Calls, Raising a 3B Model’s Success Rate on Unseen Rules From 56.3% to 88.0%

RuleMaze uses executable verifiers to test whether multimodal models can understand mazes, follow natural-language rules, and complete path planning. By separating perception, movement, and rule verification, DMP substantially outperforms standard fine-tuning, though testing remains limited to synthetic grids and the verifiers underwent manual correction.

2026-08-22

代理工具

TMI Reconstructs Task Trees from Interleaved Computer Activity, Raising Step-Description Accuracy from 30.3% to 74.9%

Task Model Induction does not compress an entire screen and input log into a single workflow. Instead, it first disentangles noncontiguous tasks, then reconstructs goal hierarchies and loop-based control flow. The resulting agent skills improved held-out accuracy from the best baseline’s 14.29% to 18.57%, but the full pipeline depends on frontier models, while raw interaction traces introduce significant privacy risks.

2026-08-22

代理工具

Brain Researcher Raises First-Step Tool Selection Rate for Brain Imaging Agents from 23.3% to 93.6%

Brain Researcher constrains neuroimaging analysis agents with a knowledge graph, typed workflows, and review rules, while preserving hypotheses, evidence, and execution provenance. Tool routing improved substantially, but the rate of successfully grounding claims in verifiable evidence remained just 22.0%, and automated review also failed to catch a directional statistical error.

2026-08-22

AI 安全

SafeBranch Pairs Safe Forks from Identical States, Raising IS-Bench Safe Success Rate from 0.031 to 0.281

SafeBranch rolls back an agent’s hazardous trajectory to a critical decision point and compares safe and unsafe actions only within the same context. The trained model requires no external critic at deployment, but results are currently limited to simulated household environments, and data construction still relies on GPT-4o judgments.

2026-08-22