WeAct
WeAct is a method for long-horizon agents that incorporates the selection of capability-specific LoRA adapters into the agent's action space. It reuses the native tool-calling interface to achieve on-demand composition of capabilities without additional routing modules, while keeping the backbone parameters unchanged, thereby avoiding catastrophic forgetting.
Timeline (last 90 days)
WeAct proposes adding capability-specific LoRA adapter selection to the agent's action space, achieving on-demand capability composition through the native tool-calling interface.
On eight deep research benchmarks, Qwen3-32B with WeAct achieves an average accuracy 27.5% higher than the strongest static LoRA merging method.
WeAct trains multiple capability-specific LoRA adapters on a frozen backbone model, with each adapter corresponding to one capability; adapter selection is modeled as an internal action of the agent, on par with calling external tools.
WeAct's internal actions reuse the model's native tool-calling interface and do not require an additional routing module or controller.
In the WeAct experiments, a new code-computation adapter was added and the existing writing adapter was updated, and neither update modified the backbone model or the other adapters.
WeAct's current validation is concentrated on deep research tasks and the Qwen3-32B base, and how the number of adapters and capability granularity scale with task complexity is not sufficiently discussed in the original text.
All Facts (6)
WeAct proposes adding capability-specific LoRA adapter selection to the agent's action space, achieving on-demand capability composition through the native tool-calling interface.
50%UnverifiedOn eight deep research benchmarks, Qwen3-32B with WeAct achieves an average accuracy 27.5% higher than the strongest static LoRA merging method.
50%UnverifiedWeAct trains multiple capability-specific LoRA adapters on a frozen backbone model, with each adapter corresponding to one capability; adapter selection is modeled as an internal action of the agent, on par with calling external tools.
50%UnverifiedWeAct's internal actions reuse the model's native tool-calling interface and do not require an additional routing module or controller.
50%UnverifiedWeAct's current validation is concentrated on deep research tasks and the Qwen3-32B base, and how the number of adapters and capability granularity scale with task complexity is not sufficiently discussed in the original text.
50%UnverifiedIn the WeAct experiments, a new code-computation adapter was added and the existing writing adapter was updated, and neither update modified the backbone model or the other adapters.
50%