analiza aiqbench
This paper proposes CABA, a post-update gate for continual learning that accepts a model update only if its divergence from the old policy stays within a set budget on held-out historical anchors. The theory is conditional on representation-coverage and Lipschitz assumptions, and the evidence is a small UCI Human Activity Recognition experiment that shows a clear retention-plasticity trade-off: zero measured forgetting costs about 4.5 points of accuracy. Readers should note that the authors explicitly frame it as a workshop-level proxy, not evidence for foundation-model or robotics performance.
Certified Behavioral Retention for Continual Adaptation with Held-Out Anchors
arXiv:2610.40740v1 Announce Type: new 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.
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.
11 pages, 1 figure, 3 tables. Accepted at the NeurIPS 2026 Workshop on Continual Learning in the Era of Foundation Models and Embodied Agents (CL4M / CL4FMAgents)
Machine Learning (cs.LG); Machine Learning (stat.ML)
arXiv:2610.40740 [cs.LG]
(or arXiv:2610.40740v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.40740 arXiv-issued DOI via DataCite (pending registration)
Text integral sub licența deschisă a publicației · 265 cuvinte
Subiecte conexe
Integrating adaptive human behavior into epidemic models with large language models
arXiv · 1 sept. 2026
EASE: Behavior-Adaptive Skill Curation for Self-Evolving Agents
arXiv · 30 sept. 2026
A System for Fast, Resilient, and Adaptable Loco-Manipulation Behaviors on Humanoid Robots
arXiv · 2 sept. 2026
A Behavior-first approach to Multi-Intention Inverse Reinforcement Learning
arXiv · 14 oct. 2026