最新模型

新模型、開放權重、多模態、語音、影像與模型架構。

AgenticASR Repeatedly Revises Transcripts With a Sliding Window, 1B Refiner Pushes Bilingual Benchmark Score to 79.95

AgenticASR separates speech recognition from text cleanup, allowing a small Refiner to rewrite a limited portion of existing output as new speech arrives, without repeatedly processing the full transcript. The code, checkpoints, and a 917-item Chinese-English benchmark are publicly available, but most of the audio is synthetically generated and scoring relies on an LLM judge.

2026-07-31

最新模型

The Chapter Organization of Synthetic Textbooks Can Itself Improve Mid-Training, Raising the Average Score by 1.02 Points Even With the Text Held Constant

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.

2026-07-31

推論系統

SemPIC Precompiles Documents into Reorderable KV Caches, Achieving Average F1 Close to Full Recomputation on Long-Context Tasks

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.

2026-07-31

推論系統

QuantWAMs Runs World Action Models Primarily at 4-bit, Targets 29% of FP16 Memory Usage for Key Modules

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.

2026-07-31

AI 安全

AISPA Audits Hidden Instructions in 88 Commercial AI Products, Finds About 40% Contain Rules That Disadvantage Users

AISPA breaks system prompts into eight dimensions of user rights and audits 3,249 instructions across 88 commercial AI products. While 98.9% of products contain at least one protective rule, only 24% cover all dimensions, and about 40% also include instructions deemed problematic.

2026-07-31

代理工具

Qwen-UI-Agent Unifies Mobile, Desktop, and Browser Operations, Extending Long-Trajectory Training Beyond 100 Steps

The Qwen team has introduced a unified GUI agent spanning mobile, desktop, web, and DeepSearch environments, with on-screen interactions and CLI commands sharing the same action space. Its technical report claims a score of 79.5% on OSWorld-Verified, but the model weights, training data, and complete trajectories have not yet been released.

2026-07-31

代理工具

OSReward Finds That Computer-Agent Judges Commonly Let Failed Trajectories Pass, While Visual Evidence Has Surprisingly Limited Impact on Verdicts

Using 1,019 human-reviewed cross-platform trajectories, OSReward tested 27 vision-language judges and found that their main error was classifying incomplete tasks as successful. Removing step-by-step text reduced accuracy by an average of 7.2 percentage points, while changing the visual inputs had a smaller effect, exposing current “visual judges’” reliance on agents’ self-reported narratives.

2026-07-31

代理工具

Tycho Lets Agents Write World Models During Interaction; Two Frontier Models Complete Every Public ARC-AGI-3 Level

Tycho translates unknown game rules into executable state-transition models, then uses validation results to decide whether to revise or bypass them. GPT-5.6 Sol and Claude Opus 5 completed all 183 levels in one public-set evaluation, though the results still depend heavily on frontier APIs, single runs, and public tasks.

2026-07-31

代理工具

Echoverse Trains Computer Agents on Resettable, Stateful Websites, Boosting 9B Model Evaluation to 67.1%

Microsoft’s Echoverse moves away from large numbers of shallow website replicas, instead building synthetic applications with database state, causal consequences, and verifiable tasks. After training on 12 environments, a 9B model improved from 36.5% to 67.1% across 14 evaluation splits, although the public release includes only four of those environments.

2026-07-31

代理工具

MCP 2026-07-28 Removes Handshake and Sessions as Remote Tool Servers Move to Stateless Requests

The new MCP places the protocol version, capabilities, and client information in every request, removing the `initialize` handshake and `Mcp-Session-Id`. This simplifies horizontal scaling but introduces a breaking change that requires coordinated upgrades to clients, servers, and gateway rules.

2026-07-30

最新模型

PerceptionBench Breaks Visual Reasoning Back Down into Ten Basic Perceptual Skills; None of 16 Multimodal Models Scores 60%

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.

2026-07-31

代理工具

Home Assistant 2026.8 Beta Natively Integrates llama.cpp and LiteLLM, Enabling Fully Self-Hosted Home Conversation Agents

Home Assistant adds conversation integrations for llama.cpp and LiteLLM, allowing users to create separate agents and instructions for different models. The update lowers the barrier to connecting local models, but it remains in beta, and end-to-end tool-calling reliability depends on the model, schema, and hardware.

2026-07-30