# When Cooperation Must Change: Semantic Strategy Unlearning in Multi-Agent Reinforcement Learning

DevFeed: [When Cooperation Must Change: Semantic Strategy Unlearning in Multi-Agent Reinforcement Learning](<https://devfeed.tech/articles/when-cooperation-must-change-semantic-strategy-unlearning-in-multi-agent-reinforcement-learning-135143.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.17917>)

Author: Wenhan Wu, Linghan Chen, Yudong Gao, Fan Yang, Honglong Chen, Kai Yin, Jiawei Jiang, Chuang Hu

Published: 2026-11-16T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Reinforcement learning](<https://devfeed.tech/topics/reinforcement-learning.md>), [Multi Agent Systems](<https://devfeed.tech/topics/multi-agent-systems.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>)

Tags: [adversarial](<https://devfeed.tech/tags/adversarial.md>), [attribution](<https://devfeed.tech/tags/attribution.md>), [auditing](<https://devfeed.tech/tags/auditing.md>), [behaviors](<https://devfeed.tech/tags/behaviors.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [cache](<https://devfeed.tech/tags/cache.md>), [completion](<https://devfeed.tech/tags/completion.md>), [content](<https://devfeed.tech/tags/content.md>), [control](<https://devfeed.tech/tags/control.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [decentralized](<https://devfeed.tech/tags/decentralized.md>), [decision-making](<https://devfeed.tech/tags/decision-making.md>), [delivery](<https://devfeed.tech/tags/delivery.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [distributed](<https://devfeed.tech/tags/distributed.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [learning](<https://devfeed.tech/tags/learning.md>), [multi-agent](<https://devfeed.tech/tags/multi-agent.md>), [reinforcement-learning](<https://devfeed.tech/tags/reinforcement-learning.md>), [semantic](<https://devfeed.tech/tags/semantic.md>), [strategy](<https://devfeed.tech/tags/strategy.md>)

## AI overview

CARVE targets obsolete or unsafe coordination conventions in cooperative multi-agent reinforcement learning while aiming to preserve useful cooperation. The paper describes the challenge as removing a relational pattern that different agents can reproduce across different histories. It combines causal attribution of influential interactions, retain-aware subspace updates, counterfactual re-coordination to learn alternatives, and adversarial audits of whether the convention can still be recovered. The abstract reports results on five cooperative MARL benchmarks: mean nominal convention-completion and worst-case elicitation rates fell by 91.4% and 83.3%, respectively, with a median utility loss of 5.9% and inference latency matching the original policies. These are abstract-level reported results; no benchmark-specific details are supplied.

## Source excerpt

arXiv:2611.17917v1 Announce Type: new Abstract: Cooperative multi-agent reinforcement learning (MARL) supports coordinated decision-making in distributed Web services, multi-robot control, and collaborative games through learned coordination conventions. When deployment requirements change, a previously useful convention may become obsolete or unsafe. For example, revised cache-sharing permissions in Web content delivery may require caches to abandon a learned division of storage and serving roles while maintaining efficient delivery. Teams must therefore remove the specified convention while preserving useful cooperation. This is challenging because the same relational pattern can be reproduced by different agents across different histories. Runtime shielding leaves the underlying coordination preference intact and can impair task progress, while existing reinforcement unlearning and trajectory-deletion methods do not directly target role-invariant conventions. This paper formalizes this problem as semantic coordination-strategy unlearning and proposes CARVE.CARVE uses causal coordination attribution to identify influential interactions and retain-aware subspace updates to limit interference with unrelated behaviors. Counterfactual re-coordination helps agents learn productive alternatives, while adversarial auditing tests whether the target convention can still be recovered. Accepted updates are merged into decentralized actors for direct deployment. Evaluation across five cooperative MARL benchmarks shows that CARVE reduces mean nominal convention-completion and worst-case elicitation rates by 91.4\% and 83.3\%, respectively.The median utility loss is 5.9\%, and measured inference latency matches that of the original policies.