AI Cosmos reports larger learning gains than LLM baselines
Original title: AI Cosmos: An Agentic Infra for Situated Learning
Two preregistered studies found stronger legal-reasoning gains and better mastery of unseen programming problems.
AI Cosmos is an agentic infrastructure for situated learning built around scenario-based participation and cumulative practice assets. Learners and role-specialized agents work in scenarios with task knowledge and, when needed, executable workbenches, while an asset-memory loop stores processes and outcomes with provenance and dependencies. Two preregistered randomized studies held the same underlying LLM across conditions. The paper reports greater legal-reasoning gains than a conventional LLM and higher mastery of unseen programming problems than an LLM-as-Teacher baseline.
- Scenarios integrate task-relevant knowledge and optional executable workbenches.
- Practice processes and outcomes are stored as reusable assets with explicit dependencies.
- Both studies were preregistered and randomized, with the underlying LLM held constant.
From the original article
1 min read at arxiv.orgLarge language models (LLMs) can improve task performance during learning, yet these gains do not necessarily reflect learners' ability to thereafter apply the knowledge and skills independently. We present AI Cosmos, an agentic infrastructure for situated learning that organizes learning around two complementary principles: scenario-based participation and cumulative assets of practice.
By AI Cosmos Team · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.