Recuris decouples memory to improve long-horizon agent reasoning
Source published ·Modelwire updated
Original coverage: arXiv cs.CL ↗·How Modelwire adds context
The development
Researchers introduce Recuris, an architecture that tackles a fundamental bottleneck in long-horizon AI agents: how to maintain task focus when execution histories grow unwieldy. The system separates working memory (tracking immediate progress) from experiential memory (storing learned skills), allowing agents to invoke capabilities based on current needs rather than full history. A meta-agent then converts execution failures into targeted, validated updates that reshape the agent's skill set. Testing across four benchmarks and ten models suggests this memory-evolution loop yields measurable gains. The work addresses a real scaling challenge for autonomous systems that must operate over extended timescales without losing coherence.
Modelwire’s AI-generated summary of coverage from arXiv cs.CL.
Modelwire analysis
ExplainerOur AI-generated reading of the wider context and the next developments to watch.
The paper doesn't just propose memory separation (that's been explored before) but specifically introduces a meta-agent that converts execution failures into validated skill updates. That feedback loop from failure to memory reshaping is the novel mechanism worth isolating.
This is largely disconnected from recent activity in the space, which has focused on scaling model size and context windows. Recuris belongs to a smaller but growing thread: internal architecture changes that help agents maintain coherence without simply throwing more parameters at the problem. The work sits between two concerns: how do long-horizon agents avoid context collapse, and how do they learn from their own mistakes in ways that stick. We haven't covered prior work in this specific intersection yet.
If Recuris shows consistent gains across the four benchmarks when tested on held-out model families (not just the ten models in the paper), that suggests the approach generalizes beyond the experimental setup. If gains disappear when the skill memory is frozen and only working memory is updated, that confirms the meta-agent loop is doing the actual work rather than just better bookkeeping.
This interpretation is generated from the summary above and available source metadata. Our methodology · Report an error
MentionsRecuris · Meta-Agent · Experiential-Working Memory · Skill Memory
How this coverage is produced
Modelwire uses AI to generate summaries and context from source headlines, snippets, and selected archive coverage. Automated checks do not verify every claim, and items are not routinely reviewed by a person before publication. Zacaria Solis operates the site. Read the linked source for the full evidence and report errors through our corrections process.
Modelwire summarizes, we don’t republish. arXiv cs.CL originally reported this story as “Recursive Experiential-Working Memory Evolution for Long-Horizon Agent Harnesses”. The full content lives on arxiv.org. If you’re a publisher and want a different summarization policy for your work, see our takedown page.