frontier models
The Agent Changes Its Weights Mid-Task
WeAct adds selection among capability-specific adapters to the action space instead of fixing one parameter configuration for an entire trajectory.
Summary
WeAct adds selection among capability-specific adapters to the action space instead of fixing one parameter configuration for an entire trajectory.
A long research task may demand browsing, calculation and writing in sequence, while updating one shared model risks forgetting an earlier skill. WeAct freezes the backbone, trains separate LoRA adapters and lets the agent select one through its normal tool-calling interface. On eight deep-research benchmarks, the authors report a 6.2 percent average accuracy gain over full fine-tuning and 27.5 percent over the strongest evaluated static adapter merge. Adding a computation adapter and updating a writing adapter left other modules unchanged. The work tests one modular strategy, not arbitrary self-modification, but it turns internal parameter choice into an explicit action.
Why it matters
WeAct adds selection among capability-specific adapters to the action space instead of fixing one parameter configuration for an entire trajectory.
Limits and context
- The work tests one modular strategy, not arbitrary self-modification, but it turns internal parameter choice into an explicit action.
Key claims
WeAct adds selection among capability-specific adapters to the action space instead of fixing one parameter configuration for an entire trajectory.
Qualification: The work tests one modular strategy, not arbitrary self-modification, but it turns internal parameter choice into an explicit action.
Evidence: source-2026-10-20-008
Sources
- arXiv preprint 2610.26847, original v1arXiv · primary research
Corrections
No corrections have been recorded for this story.