Archive
新模型、開放權重、多模態、語音、影像與模型架構。
代理工具
MIST tests not only whether models can resist incorrect prompts, but also whether they still incorporate accurate external information. Its companion method, SCOPE, reduced Qwen3-4B’s correct-to-wrong rate from 35.0% to 16.3%, though training has so far been validated on only two families of small open models.
代理工具
A new study compiles API schemas into typed Python stubs, allowing models to chain or invoke tools in parallel within a single inference. The method matches or outperforms native JSON for most models on a BFCL v4 subset, but it has so far been tested only with simulated tools that return their arguments, so the results do not establish that real-world API workflows would see the same benefits.
最新模型
GeniWorld first uses URDF to convert joint actions into pixel-aligned robot motion, then employs a causal DiT to predict how the environment will respond. Across four real-world tasks, the authors report that the overall success rate increased from 40.8% to 69.0%, although neither the code nor model weights have been released.
最新模型
A new study audits video models using executable temporal trajectories rather than checking only final counts. In the high-event-count, high-frequency regime, Gemini 3.6 Flash answered just 0.2% correctly and recovered only 18.1% of ground-truth events.
代理工具
HarnessOpt-Bench asks models to modify prompts, tools, memory, and control flow within a fixed evaluation budget, then measures actual gains on an unseen test set. Across 111 experiments, the model itself accounted for more variation than the outer coding harness, while native tools did not consistently outperform shared ones.
最新模型
The study traces the difference between Adam and low-rank recovery to the coordinate-wise preconditioner’s lack of gauge equivariance. In published experiments, two functionally equivalent initializations trained with Adam produced attention invariants that differed by 56%, though the findings have not yet been validated in large-model pretraining.
最新模型
Rather than treating teacher–student discrepancies at every token equally, DASH adjusts how far supervision can propagate based on the entire generation trajectory. The authors report consistent gains on three mathematics benchmarks and have released the training code, evaluation pipeline, and LoRA weights.
Reasoning Core 用可執行評分器、難度控制與短答案格式產生數學、規劃、形式證明及程式題。匹配訓練實驗顯示它優於三套程序資料集，但作者也在外部資料庫發現生成器與評分器不一致。
AI 安全
A new benchmark plants targeted probes—including semantic lures, parameter traps, and capability illusions—in MCP tool lists, using erroneous calls to pinpoint agents’ tool-selection weaknesses. Across 11,520 experiments, model tier did not consistently predict safety, while the ability to recover after being deceived raised task success rates from 16% to 52%.
推論系統
ReCo dynamically compresses the KV cache based on process rewards for each reasoning step, while suppressing repetitive reflection and enabling early stopping. In the authors’ experiments across three reasoning models and six benchmarks, generated tokens fell by 37% to 65%, while latency improved by 2.08× to 2.35×.
AI 基礎設施
Microsoft Azure research shows that agents repeatedly switch among model inference, tools, and orchestrators, putting the CPU back on the critical path while GPUs often sit idle. The Agora prototype can reclaim compute based on workflow patterns, but incorrectly reducing GPU allocations for parallel agents can worsen tail latency by 16×.
最新模型
Instead of simply concatenating individual libraries into long documents, OctoLong recursively gathers implementations by following AST, language-server, and package dependency relationships. The research team released models ranging from 600M to 14B parameters, with experiments showing that dependency-dense data improves code retrieval, state tracking, and API usage.