Reproduce “Why Pay for Long Reasoning When Rewrite Do Trick? Causal Effects of Reasoning History Compaction on Future Generation and Agent Cost” 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

Anton Hristov

The paper tests whether rewriting a model’s earlier reasoning into a compact state record changes its later reasoning in tool-using tasks. A compressor rewrites prior assistant text before a pending tool result is revealed; both histories then receive the same result. Across accepted rewrites, 151 of 172 continuations used fewer reasoning tokens, but task-quality preservation was not established.

Summary by alphaXiv.

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