SURL AI News
繁中English日本語

Archive

GitHub Repo

GitHub 熱門專案、開源工具、框架、runtime 與重要 release。

代理工具

ERA’s Reported 11% Edge in Retrospective Epidemic Forecasting Is Reassessed, With Data-Revision Leakage Explaining Nearly All of the Gain

A new analysis finds that Google’s ERA was backtested using season-finalized data, effectively giving it access to revisions that were unavailable to CDC models when their forecasts were made. The two achieved similar Weighted Interval Scores when the data underwent little or no revision, showing that AI forecasting benchmarks must preserve the version of the data available at each point in time.
代理工具

Cloudflare Launches Kitesurf: Lightweight Web Engine Replaces Chromium, Cutting Resource Usage for Agentic Browsing to About One-Third or Less

Kitesurf brings Rust, Wasm, and modular Web components to Cloudflare Workers, reducing browser overhead for agent workloads such as content extraction, form interaction, and screenshots. Cloudflare claims that some tasks use 3 to 7 times less CPU and memory in its tests, though website compatibility, anti-bot verification, and source-code availability remain major limitations.
AI 安全

HERALD Reveals That Search Agents Can Cite Unretrieved Passages; A Single Membership Check Reduces the Observed Attack Rate to Zero

HERALD audits search-agent rewards using question-matched counterfactual edits, finding that citations to real but never-retrieved passages can score at least as highly as the original trajectory under the baseline reward. After adding a citation–retrieval-set membership check, no successful attacks were observed across 593 eligible questions, but the reinforcement signal appeared in only 0.03% of training trajectories.
代理工具

CIPO Assigns Step Rewards by Masking Retrieved Evidence, Raising Qwen2.5-7B’s Average F1 Across Seven QA Benchmarks to 0.504

CIPO compares an agent’s action probabilities with and without access to the latest retrieval results, directly rewarding next steps influenced by external evidence. In the authors’ experiments, it outperformed the strongest baseline by 4.7 F1 points, but the additional scoring increases training costs, and the method has so far been validated on only two Qwen2.5 models.
代理工具

Synthetic Clinical Data Can Still Be 79.44% Missing Despite Passing the Utility Threshold, While Two Refinements Widen the Source-Distribution Gap

An Oracle Health team found that a Synthea-derived care-gap benchmark could remain highly sparse and templated even after passing existing proxy evaluations. Deterministic refinements improved the proportion of actionable data, but the mean gap from the operational reference population widened from 9.70 to 30.69, showing that internal realism and source fidelity cannot be collapsed into a single metric.
最新模型

u-OPSD Trains Solely Through Internal Model Voting, Boosting Qwen3-8B’s Five-Benchmark Math Average by 10.7 Points

u-OPSD uses eight reasoning trajectories from the same model to produce a majority-vote pseudo-answer, then distills token distributions conditioned on that consensus into trajectories that disagree. Qwen3-8B’s average across five benchmarks in non-thinking mode rose from 43.57 to 54.31, though incorrect consensus and the cost of repeated sampling still limit its applicability.
最新模型

BaKron Reduces Two-Sided Hessian Quantization to Cubic Complexity, With Its 8192-Square-Matrix Kernel 60× Faster Than YAQA

Using anti-diagonal parallelization and recursive divide-and-conquer, BaKron reduces total work from quartic to cubic complexity while preserving curvature information on both the input and output sides. In 2.81-bit experiments on Llama 3 8B, it improves perplexity over GPTQ, but total quantization time remains longer, and no public implementation is currently available.
最新模型

Hyper-ES First Uses Gradients to Build a Low-Dimensional Search Space, Averaging 1% Higher on Mathematical Reasoning Than GRPO-LoRA

Rather than applying random evolution directly to billions of model-weight dimensions, Hyper-ES constructs a subspace from a small number of fine-tuning directions and then searches for layer-wise model-merging coefficients. The authors report modest gains across three models and six mathematical reasoning datasets, although the method is not entirely gradient-free.
代理工具

OPERA Constrains Experimental Agents with Physics Residuals, Cutting Reward-Seeking Decisions with No Real Improvement from as High as 39.0% to 1.9%

Rather than returning only a single score, OPERA gives language models interpretable physics residuals before they select, compose, or generate optical operations. Three optical tasks show that residuals can curb specification gaming, but the public code does not include experimental instrument control or the complete statistical pipeline.
最新模型

Google Releases WeatherNext 2 Code and Weights; Full-Model Inference Requires H100-Class Memory

Google DeepMind has released the JAX implementation, pretrained weights, and a streamlined Colab-compatible version of WeatherNext 2, covering global weather and tropical cyclone forecasts up to 15 days. Researchers can now reconstruct the autoregressive forecasting workflow, but the full model still depends on ECMWF initialization data and high-end accelerators, and its outputs cannot replace official warnings.