Reproduce “You don't need long context” 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
Peter Belcak, Saurav Muralidharan, Jindong Jiang, Yonggan Fu, Wonmin Byeon, Hanrong Ye, Ryan Angilly, Ali Taghibakhshi and 4 more
The authors test whether agents need full histories or documents in active context using compact handoff states and tool-based retrieval. In long-horizon and long-input tasks these approaches preserve or improve results. For Qwen 3 Coder Next, a 16K rather than 256K context window yielded 7× more decoded tokens per second; this measures throughput, not task time, and evidence is limited.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.