Recursive Agent Harness Framework Improves Long-Context Reasoning in AI Coding Agents
A new AI architecture called the Recursive Agent Harness (RAH) has been proposed, in which a parent AI agent generates and runs scripts that spawn parallel subagents equipped with tools like code execution and file system access. RAH extends prior work on recursive language models by making the recursive unit a full agent environment rather than a bare model call. In controlled evaluations, RAH improved a coding-agent benchmark score from 71.75% to 81.36%, with further gains when using a stronger underlying model.
Researchers have introduced the Recursive Agent Harness (RAH), a framework that applies recursion at the level of full AI agent environments rather than individual model calls. Building on earlier recursive language model (RLM) research and recent production coding agents—including Anthropic's dynamic workflows—RAH allows a parent agent to generate executable scripts that spawn parallel subagent harnesses for complex, fine-grained workloads, while using structured function calls for smaller subtasks. The system provides each subagent with tools including filesystem access, code execution, and planning capabilities. In a controlled evaluation on long-context reasoning using the Oolong-Synthetic benchmark (199 samples across 13 context-length buckets up to 4 million tokens), RAH improved over the Codex coding-agent baseline from 71.75% to 81.36% when both used GPT-5 as the backbone, isolating the architectural gain to the harness design itself. Substituting the stronger Claude Sonnet 4.5 backbone pushed performance further to 89.77%. The authors frame RAH as a 'code-first' extension to model recursion, positioning it as a natural next step in scaling agentic AI systems for long-context tasks.
What's missing
The study relies on a single benchmark (Oolong-Synthetic) for evaluation, and it is unclear how RAH performs on other long-context or agentic benchmarks.
What different sources said
- arXiv cs.CLCenter
Recursive Agent Harnesses
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.