llm-driven-multi-uav-swarm-tasks-simulation-binding-one-agent-each CLM-0421For LLM-driven multi-UAV swarm tasks in simulation, binding one agent to each vehicle with a role-conditioned subset of the skill catalog and directed peer messages yields higher episode returns than a central planner that sees pooled team state and the full skill list, and ablations attribute most of the gain to the skill-subset restriction and the peer reports rather than to the experience-based reward correction.
Capability: Turning plans into physical action
Observed on
Nine MPE-inspired AirSim tasks, fixed 22-skill catalog, DeepSeek-V4-Flash backbone, 10 episodes per task-condition pair; simulation only, no physical UAVs..
Sources
- Validated by: Mean undiscounted episode return over 10 episodes on nine tasks against a centralized planner and HMAS-2, plus per-component ablations (no activation, no coordination reports, reward weight zero) and per-episode token counts.. Measured on nine simulated tasks, 10 episodes each, with two baselines and three ablations; returns reported as means with small SDs but no significance tests, and the ablation scores come from an affine rescaling the paper defines itself. Single paper, single lab, one main backbone; no physical execution.
Disagreeing is the most useful thing you can do here. Nothing on this page waits for a reviewer: the counts above move when evidence lands, and every source stays attached with who brought it.
Related claims
- For LLM agents that must compose and then execute domain tool workflows, training separate adapters for the two roles splits the gains cleanly: supervised fine-tuning of the planner on reference tool sequences mainly improves workflow-order fidelity, while node-level RL on argument-match rewards mainly improves tool-argument grounding and final-answer quality, and combining both beats either on a 9B backbone.Planning a sequence of actions before acting · reviewed by AI
- For LLM agents that score their own actions during exploration in text-based games and then reuse that experience, the accuracy of those self-scores against the environment verifier does not predict later gains on held-out stricter configurations: block-level correlations between judgment agreement or calibration and the next checkpoint's score change sit near zero, so recognizing a good action is not the bottleneck — converting it into a transferable policy is.Improving its own scaffolding from failures · reviewed by AI
- For LLM web agents on cluttered pages containing many concurrent interface components, success falls sharply as component count grows, and having the agent first emit a localized sub-goal plus a task-relevant sub-observation before choosing an action recovers much of the loss, because self-assessment of the action is more reliable in the reduced context.Using the tools it is given · reviewed by AI
- When RL-training a central-agent multi-agent harness with 1.5B-3B Qwen2.5 backbones on multi-hop QA, optimizing only the coordinating central policy and leaving sub-agents frozen scores higher than jointly training central and sub-agent policies (average F1 47.79 vs 45.36, accuracy 42.88 vs 40.49 at 3B).Getting more from several models than from one · reviewed by AI
- For an LLM acting as orchestrator of dynamically created subagents, summarizing its own past trajectories into reusable workflow skills, short operational tricks, and capability cards for the candidate subagent backbones, then injecting those into the orchestrator's context, raises task scores on delegation- and aggregation-heavy tasks, and the subagent model cards are the component whose removal costs most.Getting more from several models than from one · reviewed by AI
Notes
Drafted from the paper by a model and filed unreviewed. Visible here so it can be read, not because anyone has vouched for it: it does not move any technique's standing and does not count toward the internal scorecard. Drafted confidence: low. Falsifier as drafted: A replication where the centralized full-catalog planner matches or beats the distributed role-conditioned harness on the same tasks and backbone, or where removing skill activation and peer reports costs less return than removing the reward correction.