Newest papers

WeAct: Dynamic Weight Switching as Agent Actions

Rui Shi, Xunjian Yin, Jiacheng Sang, Zichuan Liu, Shuyan Zhou · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26847 · cited by 0 works (OpenAlex) · open-access PDF

Portuguese

Agentes de horizonte longo devem aprender capacidades heterogêneas de trajetórias interleavadas. Atualizar parâmetros compartilhados para aprender novas capacidades pode degradar as capacidades previamente adquiridas, resultando em esquecimento catastrófico. À medida que uma trajetória se desenrola, diferentes etapas podem exigir capacidades diferentes, mas a formação conjunta convencional e a fusão de adaptadores estáticos mantêm uma configuração de parâmetro fixo durante toda a execução. Introduzimos WeAct (Ações internas de nível de peso), que treina adaptadores LoRA independentes específicos de capacidade sobre uma espinha dorsal congelada e adiciona a seleção de adaptadores ao espaço de ação do agente. Essas ações internas usam a interface de chamada nativa do modelo para a composição de capacidades sob demanda em um contexto compart

Machine translation (Meta M2M100); the English original follows.

Abstract

Long-horizon agents must learn heterogeneous capabilities from interleaved trajectories. Updating shared parameters to learn new capabilities can degrade previously acquired ones, resulting in catastrophic forgetting. As a trajectory unfolds, different stages may require different capabilities, yet conventional joint training and static adapter merging maintain a fixed parameter configuration throughout execution. We introduce WeAct (Weight-level Internal Actions), which trains independent capability-specific LoRA adapters over a frozen backbone and adds adapter selection to the agent's action space. These internal actions use the model's native tool-calling interface for on-demand capability composition in a shared context. Across eight deep-research benchmarks, Qwen3-32B with WeAct improves average accuracy by 6.2% relative to full fine-tuning counterparts despite using less compute, and by 27.5% relative to the strongest static LoRA-merging method. WeAct also supports modular continual adaptation, by adding a new code computation adapter to expand the agent's capabilities and updating the existing writing adapter to improve report quality. Both updates leave the backbone and other adapters unchanged, supporting modular continual adaptation in agents.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Adapter (computing), Modular design, Computer science, Train, Computation, Distributed computing

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange