首个跨语言ML流水线生成基准，评估大模型在多语言环境下的自动化代码构建能力
arXiv:2509.22768v3 Announce Type: replace Abstract: We introduce ML2B, the first benchmark for evaluating cross-lingual task comprehension in end-to-e…
Hacker News热帖：寻找LLM在Nethack游戏中的基准测试token资源，探讨不同模型的学习表现与约束条件。
I'm thinking of benchmarking different models performance playing Nethack and learning the game under different constraints. Comments URL: https://new…
7大主流LLM在100项数据管道任务上的基准对决，看看AI能否真正胜任工程实践。
Article URL: https://apacheseatunnel.medium.com/can-ai-really-build-data-pipelines-8f8713947ad8 Comments URL: https://news.ycombinator.com/item?id=490…
量化LLM基准测试中排名的不确定性，为模型评估提供统计严谨性。
arXiv:2607.16259v1 Announce Type: new Abstract: Pretrained models are typically ranked on multi-task leaderboards to assess their effectiveness across…
从Jetson Nano到模型变体，揭示边缘LLM基准测试中易忽视的变量，避免自欺欺人。
A Jetson Nano and Ollama benchmark review appeared among DEV's top articles on 2026-07-12 . Edge inference experiments are valuable because they make …
AI代理编程的荒诞体验：如果人类做错事会被解雇，但面对AI你却想放大这种错误到上千倍。
Article URL: https://danluu.com/ai-coding/#llm-variance Comments URL: https://news.ycombinator.com/item?id=48836951 Points: 6 # Comments: 1
AI代理表现离谱却被追捧，作者自嘲式反思盲目扩大应用的荒谬现象。
Article URL: https://danluu.com/ai-coding/ Comments URL: https://news.ycombinator.com/item?id=48786334 Points: 1 # Comments: 0
无需预设数据、LLM自主生成内容的结构智能评测新方法——类比测试进入可证伪时代
arXiv:2606.21008v1 Announce Type: cross Abstract: The metanym game is a competitive word game for LLMs that measures structural intelligence against e…
揭露LLM基准测试的致命缺陷：用大模型评大模型，只会陷入循环依赖与偏见陷阱。
Article URL: https://danlevy.net/llm-evals-are-broken/ Comments URL: https://news.ycombinator.com/item?id=48572463 Points: 5 # Comments: 0
美食模拟器Overcooked启发的新型LLM工具调用基准测试，用于评估AI模型完成多步任务的能力。
Article URL: https://github.com/bassimeledath/kitchen-rush Comments URL: https://news.ycombinator.com/item?id=48551671 Points: 2 # Comments: 0
一键本地LLM基准测试工具，支持NVIDIA/AMD/Apple Silicon，极速兼容性检测
Article URL: https://github.com/JoniMartin27/inferbench Comments URL: https://news.ycombinator.com/item?id=48413247 Points: 2 # Comments: 0
用贝叶斯层次模型结合嵌入空间聚类，从根源上修正LLM基准测试的提示敏感性，让模型评估更公平可靠。
arXiv:2510.05709v2 Announce Type: replace-cross Abstract: LLM benchmarking metrics often misstate performance and uncertainty as they rely on two assu…
用项目反应理论深度剖析LLM基准测试，揭示评估偏差与模型能力真实度量。
arXiv:2605.30504v1 Announce Type: new Abstract: LLM benchmark labels are frozen at release and silently propagated into downstream benchmarks, errors …
点云数据能否让大模型更懂空间？这项偏误控制研究揭开了真相。
arXiv:2504.04540v2 Announce Type: replace-cross Abstract: 3D Large Language Models (LLMs) leveraging spatial information in point clouds for 3D spatia…
仅用5个基准就能高精度预测LLM在其余52个上的表现，大幅降低评估成本。
Article URL: https://alex.smola.org/posts/34-benchmark-selection/ Comments URL: https://news.ycombinator.com/item?id=48275201 Points: 3 # Comments: 0
首个专门评测LLM生成SVG宝可梦能力的基准测试，看各大模型在图形生成上的真实排名。
Article URL: https://svg-bench.fenx.work/ Comments URL: https://news.ycombinator.com/item?id=48138312 Points: 2 # Comments: 0