SURL AI News
繁中English日本語

Archive

代理工具

Agent、MCP、coding agent、工作流自動化與評測。

最新模型

SGLang 0.5.16 Dynamically Adjusts Speculative Verification Length Based on Confidence and Overhauls Caching for Hybrid Architectures

The new SGLang release introduces DSpark, which changes the verification window for each round based on the draft model’s confidence, and makes UnifiedRadixTree the default cache for sliding-window, Mamba, and sparse-attention models. The release also removes several quantization backends and changes the rollout transport format, so compatibility must be checked before upgrading existing services.
最新模型

LLaDA 2.2 Lets Diffusion Models Add and Delete Tokens Mid-Generation, Targeting Iteratively Correctable Agent Loops

inclusionAI has open-sourced the 100B-class LLaDA2.2-flash, using Levenshtein Editing to let parallel decoding insert, delete, and replace content. The model outperforms the company’s own autoregressive baseline in some agent benchmarks and throughput tests, but still trails in code repair and function-calling accuracy, while serving support remains incomplete.
代理工具

FluxBench Puts Chip Agents Through the Full RTL-to-GDS Flow, With Performance Gaps of Up to 86% on the Same Model Architecture

FluxBench uses standardized prompts, tools, and technology libraries to test agents across the full flow from RTL generation through synthesis, placement and routing, and ECO, rather than evaluating only one-shot code output. The authors report that different agent frameworks using the same base model can exhibit a performance gap of 86.27%, although the commercial EDA experiments are limited in scope and the paper was previously withdrawn and resubmitted.
代理工具

Dynamic Forecasting Leaderboard Reveals the Advantage of Specialized Agents: Frontier General-Purpose Models Do Not Dominate ForecastBench

ForecastBench’s preliminary leaderboard, updated July 27, shows specialized Torchcast AI agents taking all three top spots, with the highest Brier Index reaching 65.6. The GPT-5.5 search agent and the median superforecaster score fall within a similar range, but differences in question counts, submission timing, and confidence intervals limit direct ranking comparisons.
推論系統

AMD Open-Sources Hyperloom to Automatically Rewrite and Validate GPU Inference Optimizations Through a Multi-Agent Closed Loop

AMD has released ROCm Hyperloom 1.0.0a1, enabling agents to handle everything from performance profiling to kernel and host-code modifications, then decide whether to retain changes based on correctness and speed tests. The initial release supports only three Instinct GPUs, and the claim of reducing optimization time “from weeks to hours” remains AMD’s own; adopters will need to reproduce the results with their own models.
AI 研究與訓練

VIGOR Dynamically Allocates RLVR Rollouts Based on Reward Variance, Cutting Sampling for Mathematical Reasoning Training by Up to 2.3×

A new study replaces the fixed number of reasoning trajectories generated for every problem with a strategy that concentrates additional compute on samples exhibiting greater reward disagreement. The authors report higher rollout efficiency on mathematics and coding tasks, but have not released their code, while actual GPU time and scalability to larger models remain to be verified.
代理工具

Antigravity CLI 1.1.7 Stops Disabled Plugins from Running Hooks and Adds Full Context to Compound Command Approvals

Google has fixed an issue in Antigravity CLI where hooks, rules, and other custom components continued to load after a plugin was disabled, preventing broken or untrusted scripts from continuing to interfere with agent workflows. The new version also displays compound shell commands in full when approval is required, but relaxes OAuth issuer validation for compatibility with non-standard MCP providers—a change that deployers should review separately.
最新模型

AMD Opens the Full Instella-MoE Training Pipeline, With the 16B Model Activating Only 2.8B Parameters per Token

AMD has released six Instella-MoE checkpoints spanning pretraining, mid-training, long-context extension, SFT, DPO, and reinforcement learning, along with the training frameworks, data recipes, and inference code. The model uses Gated MLA and FarSkip-Collective to reduce attention and cross-GPU communication costs, but a formal technical report and independent evaluations are still pending.
最新模型

Photon-1 Learns to Operate Computers from Unlabeled Screen Videos, Replacing Action Labels with Latent-State Prediction

Induction Labs has introduced Photon-1, a 106B-A5B sparse MoE model that learns during pretraining solely by watching screen videos, without requiring mouse or keyboard action labels. The model first predicts the next latent screen state, then generates the actions needed to reach it, though current evidence of its performance comes primarily from the team’s internal evaluations.
代理工具

Cursor Router Assigns Coding Models Per Request, While Cost Mode Requires Retaining Grok 4.5 as a Routing Option

Cursor has transformed Auto mode into a Router that classifies each request and selects a model, allowing enterprises to switch among Intelligence, Balance, and Cost strategies. The company claims online testing shows savings of 60%, but hidden-by-default model selections and mandatory routing options for certain modes make auditability a key deployment concern.