Archive
新模型、開放權重、多模態、語音、影像與模型架構。
推論系統
PhyAI uses model adapters to preserve each architecture’s conditioning, solver, and action semantics while sharing graph execution, kernels, caching, and multi-GPU serving. Public benchmarks cover π0, π0.5, GR00T N1.7, Cosmos3, and MiniCPM-Robot, although some comparisons do not use exactly the same numerical precision.
代理工具
Ai2 built a replayable evaluation from 462 real math tutoring transcripts, asking models to choose between providing scaffolding and requiring deeper student reasoning. Explicitly prompting for this tradeoff improves the performance of all seven models, but simulated students and model-based grading still cannot demonstrate real learning outcomes.
應用研究
Rather than treating each board position as an independent sample, Otter incorporates the previous 20 moves and time-control information. The model outperformed Maia-2 on the authors’ Lichess rapid test, but was not directly compared with the published Maia-3 using the same data split.
代理工具
Successful demonstrations may not apply to states reached after an agent takes a detour, and unconditionally providing the full trajectory can even reduce the probability of the correct action. SMRC-SD adds a self-distillation signal only when the state is compatible with a demonstration step, outperforming the full-path baseline on both ALFWorld and WebShop.
代理工具
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.
最新模型
LFM2.5-2.6B uses a hybrid architecture combining short convolutions with GQA to bring agentic reasoning, tool use, and long-context capabilities to smartphones and personal computers. Liquid AI reports decoding speeds of 220 tokens/s on an M5 Max, though the performance and agent benchmarks are based primarily on the vendor’s own evaluations.
代理工具
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.
最新模型
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
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
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
FinEvo-Bench interleaves related but distinct financial cases to test whether agents can turn prior feedback into reusable memories or skills for subsequent tasks. All four agent harnesses outperformed paired controls that reset state after each task, though the results used only a single model backbone.
代理工具
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.