🤖 AI Summary
This work investigates how coordination content—such as shared states or tool outputs—in multi-agent or memory-augmented large language models consumes tokens within a fixed context budget, thereby displacing task instructions and evidence and degrading performance. To quantify this “task budget displacement” effect, the authors introduce the Roundtable Context Window Test (RCWT), which systematically controls total token budget, content ordering, task type, and evaluation metrics to disentangle the impact of coordination content volume from competition for contextual resources. Through controlled prompting, context window scaling, ablation studies, and experiments across multiple models—including GPT-4.1-mini, Claude Haiku 4.5, and Gemini 2.5 Flash—they find that performance sharply declines when fewer than a few hundred tokens remain for task evidence. However, if the evidence is fully preserved, models can achieve perfect task completion even when coordination content occupies up to 95% of the context window.
📝 Abstract
Multi-agent and memory-augmented LLM systems often place coordination content, shared state, prior discussion, tool outputs, summaries, and role instructions, inside the same finite prompt used for the current task. This creates a practical allocation problem: every token spent on coordination is unavailable to task instructions or evidence when a call is assembled under a fixed context budget. We introduce the Roundtable Context Window Test (RCWT), a controlled protocol for measuring this task-budget displacement effect. RCWT varies coordination content while controlling total budget, position order, task family, and scoring. In the main context-dependent recall task at $W=4096$, three commercial models remain near baseline through moderate overhead and then degrade sharply once residual reference evidence falls to a few hundred tokens. Window-scaling summaries are consistent with a task-specific residual-budget interpretation rather than a fixed percentage threshold, but we treat this as descriptive evidence rather than a universal law. To test whether the fixed-budget cliff persists when task evidence remains intact, we add an intact-task ablation: the full task/reference block is kept present while coordination tokens increase by expanding total prompt length. In that setting, all tested calls return every scored field correctly across GPT-4.1-mini, Claude Haiku 4.5, and Gemini 2.5 Flash up to a 95\% coordination ratio. This ablation narrows the claim: the main RCWT cliff is best read as task-budget displacement, not as proof that coordination volume alone causes semantic interference in the original open-ended task. RCWT is therefore a measurement primitive for context-allocation budgeting, not a complete theory of multi-agent benefit or session-level coordination.