SURL AI News
繁中English日本語

Archive

硬體

AI 晶片、GPU、資料中心、邊緣裝置與硬體安全。

最新模型

Nemotron Wins IMO Gold With a Three-Model Generate–Verify–Revise Pipeline, but the Full Competition Run Consumed 2.31B Tokens

NVIDIA has released the weights for two Nemotron 3 Ultra math specialists, their training data, and a resumable natural-language proof pipeline. The system’s official score was 30 out of 42. The results also expose a shared blind spot among model-based verifiers: internal evaluations overestimated performance on two problems, while high precision came at the cost of extensive false rejections and substantial compute.
GitHub Repo

Edge0 Streams MoE Experts from SSD, Requiring Just 2.9 GiB of Active Memory for a 35B Model at Short Context Lengths

The new open-source framework uses predictive routing to preload experts for the next step and Recover-LoRA to compensate for int4 quantization loss, so the entire 23GB checkpoint does not need to remain resident in memory. Official M4 Pro tests measured 14.9–17.7 tokens per second, but the framework currently supports only Apple Silicon, and its memory and quality results have yet to be independently reproduced.
推論系統

NVIDIA Dynamo Splits Multimodal Inference into Three Stages, Cutting Time to First Token by Up to 42.2% Under Mixed Traffic

Dynamo can now assign visual encoding, LLM prefill, and decoding to separate workers and different GPU tiers, preventing image requests from blocking text-only traffic. Although NVIDIA’s tests show improvements of up to 5× in time to first token and 7× in end-to-end latency, long outputs or large dense models may offset the benefits of disaggregation.
推論系統

PELM Jointly Tunes DVFS and Speculative Decoding, Cutting On-Device LLM Energy Consumption by Up to 52.4%

PELM goes beyond adjusting processor frequency: it dynamically selects speculative decoding configurations and verification depth, jointly controlling hardware and model workload under thermal constraints. The paper reports speedups of up to 23.1% and energy savings of up to 52.4%, but these peak results cannot be directly extrapolated to smartphones or other models.
最新模型

Cognition Launches SWE-2: Training Multiple Reasoning-Cost Levels in a Single Reinforcement Learning Run

Built on the 2.8T-parameter Kimi K3 base model, SWE-2 improves coding-agent efficiency through cost penalties, low-precision rollouts, and a draft model updated in sync with training. Official benchmarks show performance approaching that of frontier closed-source models, but the results were largely produced by Cognition itself, and neither a standalone API nor model weights are currently available.
具身 AI 與機器人

FolDeX Evaluates Long-Horizon Garment Folding With More Than 2,000 Hours of Real-World Robot Data

FolDeX turns garment folding into a real-world benchmark spanning more than 20 tasks and over 10 robot embodiments, specifically testing whether heterogeneous data can be reused across tasks, environments, and hardware. The team also offers remote evaluation on physical robots, but the data license, full download process, and first set of comparable baselines remain to be confirmed.
最新模型

Gander Open-Sources a Full-Duplex Multimodal Agent, Dividing Parallel Conversation and Long-Running Tasks Between a “Cerebellum” and “Cerebrum”

Gander brings streaming audiovisual conversation, interruption control, and asynchronous agent execution onto a single timeline, allowing the front-end model to continue interacting with users while the back-end agent works. The initial release includes Apache 2.0-licensed code and model weights, but the full dataset has yet to be released, and the official deployment configuration requires three GPUs.
GitHub Repo

MiniCPM5-2B Opens Training Data and On-Device Deployment Formats, but Quantized-Version Sampling Defaults Draw Scrutiny

OpenBMB has released a 2B dense model with a 131K-token context window and tool-calling capabilities, along with staged checkpoints, training data, and multiple quantization formats. Third-party evaluations affirm its competitiveness within its class, but community testing suggests the official sampling settings may cause GGUF quantizations to fall into repetitive generation loops.