COCORELI: Cooperative, Compositional Reconstitution \& Execution of Language Instructions
This addresses the problem of unreliable LLM behavior in complex collaborative tasks for AI systems requiring robust instruction execution.
The paper tackles limitations of large language models in complex instruction following, hallucination, and spatial reasoning by introducing COCORELI, a hybrid agent framework that integrates medium-sized LLMs with novel abstraction mechanisms and a discourse module. Experiments show COCORELI outperforms single-LLM CoT and agentic LLM systems using larger LLMs, largely avoids hallucinations, identifies missing information, asks for clarifications, and updates learned objects.
We present COCORELI, a hybrid agent framework designed to tackle the limitations of large language models (LLMs) in tasks requiring: following complex instructions, minimizing hallucination, and spatial reasoning. COCORELI integrates medium-sized LLM agents with novel abstraction mechanisms and a discourse module to parse instructions to in-context learn dynamic, high-level representations of the environment. Experiments on natural collaborative construction tasks show that COCORELI outperforms single-LLM CoT and agentic LLM systems, all using larger LLMs. It manages to largely avoid hallucinations, identify missing information, ask for clarifications, and update its learned objects. COCORELI's abstraction abilities extend beyond ENVIRONMENT, as shown in the ToolBench API completion task.