Routed Continual Learning
Michael Philip, Mahesh Balakrishnan, David Geraghty, David Mao, Vidhya Venkat and Ilya Mironov (FAIR at Meta) treat agent continual learning as a routing problem that chooses both a memory system and a state cluster for each run.
Ask this paper
Architecture. Past trajectories are grouped into clusters, and each cluster keeps state in one memory system from a fixed library. An LLM router reuses a cluster or creates a new one before each run.
Context Memory Adapter. A shared runtime maps prior experience into a memory schema and supplies learned state through the agent's context.
Global history fails on mixed work. Global in-context learning, the strongest system on isolated CL-Bench tasks, gains a mean of only +0.015 in aggregated normalized reward over the stateless condition when tasks are interleaved.
Reward and cost. On 301 CL-Bench and 97 banking τ³-Bench instances across four language models, Routed CL raises reward over Global ICL by 15 to 30% of the maximum attainable reward and cuts estimated API cost by 65 to 72%.
Untested conditions. Runs are sequential. The authors do not evaluate simultaneous updates, stale reads or cross-model reuse of cluster state.
Abstract
Agents deployed at scale repeatedly encounter related problems, and it is natural to let them learn from each other's experience rather than starting over each time. Unfortunately, there is no single way to retain that experience that works well across settings, and no way to know in advance which future problems a given experience will help with. A global history preserves every detail but becomes expensive and mixes together experience that may not belong together; compressing that history into one fixed representation limits the cost but can throw away valuable details. We introduce Routed Continual Learning (Routed CL), an architecture that decides how to partition and retain experience. Past trajectories are grouped into clusters, each of which maintains state using one memory system from a fixed library. Before each run, an LLM-based router either reuses an existing cluster or creates a new one, basing its decision on the problem statement, memory system descriptions, and cluster summaries. We also introduce Context Memory Adapter, which mediates between the memory system and the agentic loop. We evaluate the architecture on a heterogeneous stream that interleaves 301 instances from the six CL-Bench tasks with 97 banking instances from $τ^3$-Bench. Across four language models and five matched rollouts, Routed Continual Learning raises normalized reward over a single global in-context learning (ICL) history by 15-30% of the maximum attainable normalized reward, achieves higher mean normalized reward than ICL-only clustering, and reduces estimated API cost by 65-72% relative to Global ICL. These results support approaching continual learning as a routing problem that chooses both a learning mechanism and a state cluster rather than committing to one representation globally.