Language agents improve skills across tasks without forgetting
Semantic Projection for Continual Self-Evolution of Language Agents
Artificial Intelligence
Summary
Language agents need to keep getting better at new skills without losing old ones. The authors identify a problem where improving on new tasks makes the agent forget earlier skills. They developed a method called Semantic-Scope Projected Evolution (SSPE) that helps the agent update its skills by balancing gains and losses across tasks. Tests show SSPE helps the agent keep performing well across many different tasks and even when switching to new setups.
What this means in practice
- •For ai platform engineers: Maintain and update language agent skills across changing task requirements without losing earlier abilities.
- •For chatbot developers: Create chatbots that improve their conversational skills over time while retaining older knowledge for diverse interactions.
Authors
Ziyu Liu, Jun Chen, Lixu Wang
Abstract
Language-model agents increasingly rely on persistent natural-language skills to adapt beyond their frozen model parameters. When a shared skill is repeatedly revised from a non-stationary, heterogeneous task stream, however, improvements for new tasks can overwrite procedures needed for earlier ones. In continual learning, Orthogonal Gradient Descent (OGD) addresses analogous interference by projecting a new-task gradient onto a subspace that locally preserves prior predictions. Natural-language skill revisions, however, have neither gradients nor a canonical vector space in which such a projection can be performed. We introduce \emph{Semantic-Scope Projected Evolution} (SSPE), which transfers the functional principle of gradient projection from parameter space to behavior space. SSPE treats an unconstrained skill revision as a proposed update, identifies acquired capabilities with which it may interfere, and uses the observed gains and regressions to construct a compatible revision rather than merely rejecting the update. This enables one shared skill to evolve across latent and recurring task contexts without exposing semantic domain identities to the evolution model. Across controlled synthetic streams and heterogeneous real-agent benchmarks, SSPE improves final cross-domain competence and mitigates forgetting relative to strong skill-evolution baselines. The evolved skill also retains the strongest average performance after transfer to a different executor model. These results establish semantic projection as a promising principle for stable and adaptive self evolution of language agents.