Yingxin Jia⊕

About Me

I am a Research Engineer at Meta, building and evaluating AI agents that millions of people and businesses use every day. I have worked on personal and business AI assistants, as well as coding and computer-use agents.

My research asks how we can rigorously measure what AI agents can really do. I contribute to multiple public LLM benchmarks, including work published in Nature and cited 1,000+ times.

Publications

When Does Experience Memory Help Closed-Loop LLM Driving Agents? A Controlled Evaluation with Resampled Warmup Pools, Token-Matched Fillers, and Audited Baselines

Yingxin Jia, Xuanyi Li
Asian Conference on Machine Learning (ACML 2026)
Oral Presentation
PDF arXiv

Give the Model a Finger: Developmentally Aligned Evaluation of Counting in Vision-Language Models

Yingxin Jia, Hui Geng, Wenxin Yang, Mohan Zhang
NeurIPS 2026 BabyVLM Workshop
Oral Presentation
PDF arXiv

A Benchmark of Expert-Level Academic Questions to Assess AI Capabilities (Humanity's Last Exam)

Center for AI Safety, Scale AI & HLE Contributors Consortium (including Yingxin Jia)
Nature, Volume 649, Issue 8099, pp. 1139–1146, 2026
Nature arXiv Benchmark