ISE: New Dataset and Method for Training Multi-Turn Operating System Agents
A team of researchers has introduced ISE (Intent → Simulate → Execute), a three-stage pipeline for synthesizing high-quality training data for AI agents that operate computer operating systems. The framework generates structured user intents, simulates multi-turn conversations grounded in real execution outcomes, and runs tool calls in live isolated OS environments. Fine-tuning a Qwen3-8B model on the resulting dataset nearly doubled benchmark performance, outperforming both zero-shot GPT-4o and the four-times-larger Qwen3-32B base model.
Published as a preprint on arXiv, the ISE paper addresses a recognized gap in training data for OS agents: existing datasets lack structured user intents, realistic multi-turn task delegation, and grounded tool execution simultaneously. Stage 1 of the pipeline uses a four-dimensional framework (Persona, Domain, Task, Complexity) to generate approximately 50,000 intents, yielding 43,956 unique intents after deduplication with a Vendi Score of 61.57 indicating strong diversity. Stage 2 employs a role-locked user simulator that ties each conversational turn to actual execution results, producing 23,132 complete trajectories averaging over 68 total dialogue turns each. Stage 3 executes every tool call inside a live, sandboxed OS workspace, capturing authentic failure-and-recovery dynamics rather than synthetic responses. Fine-tuning Qwen3-8B on the resulting ISETrace dataset improved ClawEval pass@1 from 19.3% to 37.7%, surpassing zero-shot GPT-4o and the larger Qwen3-32B. An ablation study confirmed that the multi-turn simulation stage accounts for a large share of the performance gain. All code and data have been released publicly.
What's missing
ClawEval is used as the sole benchmark; generalization to other OS-agent evaluation suites is not demonstrated. The computational cost of running Stage 3 live OS execution at scale is not discussed, which is relevant to reproducibility. The paper is a preprint and has not yet undergone peer review.
What different sources said
- arXiv cs.AICenter
ISE: An Execution-Grounded Recipe for Multi-Turn OS-Agent Trajectories
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.