Certified Behavioral Retention for Continual Adaptation with Held-Out Anchors
Organizations: Notre Dame College
Abstract
Continual adaptation requires deciding not only how to update a model, but whether to commit an update without exceeding a declared budget for changes to protected behavior. We introduce Certified Anchor-Based Adaptation (CABA), a candidate-agnostic post-update gate that reserves historical observations from candidate optimization, compares old and candidate policies on frozen representations, and selects the largest tested interpolation coefficient whose worst-anchor forward KL divergence remains within budget. The finite check requires neither action labels nor complete trajectories. We develop a theory-led certification framework that separates measured anchor retention from conditional off-anchor guarantees. Under explicit representation-coverage and Lipschitz assumptions, an accepted update satisfies a uniform behavioral-divergence bound on protected support. The analysis gives regularity bounds for affine softmax heads, probabilistic coverage and compression guarantees, interpolation monotonicity conditions, grid-resolution bounds, and trajectory-level and sequential-composition guarantees. In a controlled UCI Human Activity Recognition study used as a non-stationary sensor-policy proxy, CABA reduces mean forgetting from 1.01 percent for unconstrained adaptation to 0.00 percent at budget epsilon = 0.05, while reducing final average accuracy from 93.04 percent to 88.48 percent. The experiment illustrates a retention-plasticity trade-off and does not establish foundation-model or robotic performance. CABA provides an auditable retention check for candidate generators exposing comparable policy distributions and a declared update path, while deployment also requires separate evidence of current-task utility.
Figures & tables
| Approach | Retained information | Role in update | Primary retention evidence |
|---|---|---|---|
| GEM / replay | Labeled past examples | Candidate training | Gradient or rehearsal performance |
| Hindsight anchors | Learned labeled anchors | Candidate training | Empirical forgetting reduction |
| TRPO / LCPO | Current or old policy states | Candidate optimization | Trust-region or method-specific analysis |
| STABLE | Anchor questions and metrics | Post-candidate LoRA gate | Empirical rescale/reject behavior |
| Memory Anchors | Labeled conflict experiences | Replay selection | Robot-policy retention |
| CABA | Held-out observations or representations | Post-candidate commit gate | Worst-anchor test plus conditional certificate |
| Method | Final average accuracy | Forgetting (pp) |
|---|---|---|
| Finetune | ||
| Replay | ||
| CABA ( ) | ||
| Frozen |
Appendix figures & tables1 asset
Supplementary material from the paper’s appendix.
Appendix
| Field | Required record |
|---|---|
| Data stream | Dataset version, official split, subject-to-group mapping, and exclusions |
| Candidate training | Optimizer, learning rate, batch size, epochs, initialization, and seed |
| Anchor split | Split fraction, sampling seed, capacity, replacement rule, and representation version |
| Gate | Divergence direction, budget, grid, numerical precision, and batching |
| Decision | Accepted coefficient, maximum anchor score, worst-anchor identifier, utility-key result |
| Models | Hashes for old, candidate, committed parameters, and frozen backbone |
Continual Learning
Continual Learning studies how models acquire knowledge from changing data streams while retaining or selectively revising what they learned before.
See all 323 papers