Reproduce “TeleChat to Xing4: Advancing Frontier Agentic Models on Domestic Chips and Frameworks” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Xingchen AGI Lab, Baotian Hu, Bingkai Yang, Boyang Wang, Changzai Pan, Chao Lyu, Chao Wang, Chen Li and 131 more

Xing 4 reports engineering approaches for scaling language models across long-context inference, distributed training, and multi-turn agent workloads. It combines compressed attention with sparse retrieval and shared indices, alongside communication-aware training schedules and trajectory management. In a simulated 285B-parameter attention study, its design reaches 4.68× normalized prefill throughput over MLA alone at 256K context; this is not a production-serving measurement.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.