Archive
GitHub 熱門專案、開源工具、框架、runtime 與重要 release。
Task-CoEvolve dynamically selects validation tasks based on historical disagreement among candidate harnesses, then estimates overall performance using sampling probabilities. On Terminal-Bench 2.1, it approaches full-set search performance, but the paper uses the same tasks for search and evaluation, and the code has not yet been released.
推論系統
A new method lets models adapt directly to KV eviction strategies such as H2O during fine-tuning, raising accuracy on some 128K tasks from roughly 20%–64% to over 90%. However, it does not consistently lead on natural question-answering benchmarks, and the open-source implementation remains positioned as a research tool.
推論系統
ReCache rewrites attention and positional encoding so that individual tools or skills can reuse their KV caches across requests. Qwen3 experiments show up to a 3.655× improvement in time to first token, but the method requires model fine-tuning and is not yet a drop-in inference plugin.
推論系統
Teams from Tencent WeChat and the Chinese Academy of Sciences rewrote the Hopper sparse-attention kernel, adding FP8, paged KV cache, continuous batching, and mean compensation for pruned blocks. In H20 tests, kernel-level speedups reached 47.26×, but end-to-end gains were smaller, and the current integration depends on the SGLang fork bundled with the project.
AI 可解釋性研究
The new method treats the distribution of reasoning outcomes as a piecewise-smooth sequence, uses PELT to identify genuine decision points, and applies Dirichlet kernel pooling to suppress sampling noise. Experiments on two 8B models show that it can substantially reduce resampling, though the findings currently cover only tinyMMLU multiple-choice questions.
GitHub Repo
Teams including Google Cloud AI Research apply the agent harness concept to the environment layer, using composable interfaces to alter initial states, visible information, and task flows. Open-source experiments span five benchmarks and deliver gains of up to 9 percentage points, although reproducing the full results still requires multiple external environments and commercial model APIs.
代理工具
Nsight AI can now connect live CUDA documentation to MCP clients such as Codex and Claude, or deploy a complete coding-assistance backend locally. The open-source Blueprint combines generation, completion, embedding, and reranking models, but the reference configuration requires a Hopper-class GPU and at least 200GB of storage.
代理工具
Anthropic’s updated tools can perform multiple interface actions in a single model turn while combining page structure, versioned skills, and persistent files. Existing beta integrations will continue to work, but upgrading the computer use tool changes the request format, and engineering teams must still manage browser isolation and prompt-injection risks themselves.
AI 安全
PolicyGuide does more than block policy-violating actions before tool calls: it tracks incomplete policy steps during every user turn. It improved reliable success rates across three τ²-bench domains, but adds one LLM verification call per turn, and some procedural-compliance metrics were designed by the authors themselves.
推論系統
Google DeepMind formulates the decision of whether to run costly model evaluations as a “Pandora’s box” search problem, deciding sequentially whether an additional assessment is worthwhile based on its value of information. It achieved the lowest average routing regret plus evaluation cost across all three tests, although its effectiveness depends on calibrated signal distributions and cost settings.
代理工具
One week after releasing Mojo 1.0, Modular has made the KGEN compiler, tools, and standard library available under Apache 2.0 with the LLVM exception. The source code can now be inspected and built independently, though the compiler is not yet accepting external contributions, and MAX remains under a separate license.
最新模型
A new study turns NoThink, Short, and Long into the model-generated first token, allowing it to allocate a reasoning budget for each problem without a separate router. On MATH-500, the method maintains accuracy close to the base model while cutting tokens by 76% on the easier GSM8K benchmark.