Google的新研究发现，给模型喂对TPU文档，比堆参数管用。JAXBench测50个真实负载，结果很实在。
Google、哈佛、UC Berkeley联合发布JAXBench，包含50个JAX工作负载，基于Llama-3.1、DeepSeek-V3、Mixtral、Mamba-2、AlphaFold2等真实架构。通过Gemini 3 Flash结合TPU文档优化，正确率从5.8%提升到37.3%，解决48个基准，速度提升1.28倍。束搜索进一步提速至1.36倍。研究表明正确率是文档问题，速度是搜索问题。
Great technical paper from Google. Great read on why context beats scale for agents working against...
Great technical paper from Google. Great read on why context beats scale for agents working against unfamiliar APIs. (bookmark it) GPU kernel optimization has KernelBench to hillclimb on. TPUs had nothing, and the Pallas DSL is documented thinly enough that models mostly guess. New research from Google, Harvard, and UC Berkeley introduces JAXBench, 50 JAX workloads built from real MaxText architectures like Llama-3.1, DeepSeek-V3, Mixtral, Mamba-2, and AlphaFold2. Eight operators ship with hand-tuned Pallas kernels from Tokamax, so agent output gets measured against expert work instead of a naive baseline. With Gemini 3 Flash, conditioning on curated TPU documentation lifts per-sample correctness from 5.8% to 37.3%, solving 48 of 50 benchmarks at a 1.28x geomean speedup. Beam search then pushes it to 1.36x. Correctness turned out to be a documentation problem and speed turned out to be a search problem. Paper: arxiv.org/abs/2607.20466 Learn to build effective AI agents in our academy: academy.dair.ai 💬 1 🔄 0 ❤️ 1 👀 670 📊 1 ⚡