aifollow.news
Back Rohan Paul
Rohan Paul· @rohanpaul_ai · X· · Original publication time AI score54

Singularity-Bench Tests Whether AI Agents Can Build Training and Inference SystemsMachine translation

AI introduction

Singularity-Bench gives agents six end-to-end tasks, including pre-training a model and building a serving engine, and compares their results with existing systems on the same GPUs. The source reports a Singularity-Index of 1.25 for Opus 5.5 versus an average of 0.58 for 10 other agents; its serving engine earned USD 11.52 per GPU-hour, compared with USD 8.18 for SGLang.

Original post · AI translation

@rohanpaul_ai
Translation · refer to the original

– https://t.co/bHzX4lEcbW

标题:“Singularity-Bench:智能体能构建智能体吗?”

ReplyRohan Paul@rohanpaul_ai
Translation
卡内基梅隆大学的一篇新论文发现,Opus 5.5 + Claude Code 构建了预训练流水线、推理引擎和集群编排器;在相同的 GPU 上,它们的表现达到或超过了 SGLang 和 NVIDIA Dynamo 等生产环境工具。 Singularity-Bench 为智能体提供 6 项端到端任务,从预训练模型到构建服务引擎。每项结果都在相同的 GPU 上,与 SGLang 或 Claude Code 等生产环境系统进行比较并评分。 Opus 5.5 的 Singularity-Index 得分为 1.25,其中 1 代表人类基准水平,而其他 10 个智能体的平均得分为 0.58。其服务引擎每 GPU 小时获得 11.52 美元,相比之下,SGLang 为 8.18 美元。
Original

New Carnegie Mellon paper finds that Opus 5.5 + Claude Code built a pre-training pipeline, an inference engine, and a cluster orchestrator that matched or beat production tools like SGLang and NVIDIA Dynamo on the same GPUs. Singularity-Bench gives agents 6 end-to-end jobs, from pre-training a model to building a serving engine. Each result is scored against a production system, such as SGLang or Claude Code, on the same GPUs. Opus 5.5 scored a Singularity-Index of 1.25, where 1 equals the human baselines, while the other 10 agents averaged 0.58. Its serving engine earned 11.52 USD per GPU-hour against 8.18 for SGLang.

View this post on X ↗

Source: Rohan Paul · x.com

Research
Found an error? Send a correction