OpenSkill: Framework Enables LLM Agents to Self-Improve Without Labeled Training Data
Researchers have proposed OpenSkill, a framework that allows large language model agents to build skills and verification signals from scratch using only open-world resources such as documentation, repositories, and the web. Existing self-evolving agent approaches require curated skills, successful trajectories, or verifier signals, which are often unavailable in real-world deployments. OpenSkill addresses this gap by enabling agents to adapt after deployment with no target-task supervision, outperforming baselines on three benchmarks.
A team of researchers has introduced OpenSkill, a framework designed to enable LLM-based agents to undergo self-evolution in open-world environments without any target-task supervision. Unlike prior approaches that depend on curated skills, successful trajectories, or external verifier signals, OpenSkill bootstraps its own learning loop by acquiring grounded knowledge and verification anchors from publicly available sources such as documentation, code repositories, and the web. These resources are synthesized into transferable skills, which are then refined through self-built virtual tasks grounded in the acquired anchors rather than in ground-truth answers. Evaluated across three benchmarks and two target agent architectures, OpenSkill achieved the best automated pass rates while adhering to the no-supervision constraint. Notably, the skills learned by OpenSkill transfer across different models without requiring model-specific adaptation, and the self-built verifier aligns with ground-truth outcomes despite never having accessed them. The paper, submitted to arXiv on June 4, 2026, spans 20 pages with 4 figures and 8 tables, and accompanying code has been made publicly available.
What's missing
The study does not report results on tasks requiring real-time or highly dynamic world knowledge, leaving open questions about performance in rapidly changing environments. It is also unclear how OpenSkill scales with model size or how it performs when open-world documentation is sparse, low-quality, or adversarial. The paper has not yet undergone formal peer review, as it is a preprint.
What different sources said
- arXiv cs.CLCenter
OpenSkill: Open-World Self-Evolution for LLM Agents
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.