Formalizes continual model routing (CMR), releases CMRBench with over 2000 models, and presents CARvE which outperforms retrieval, fine-tuning and adapter-merging baselines on model/family/domain accuracy.
van de Ven, Tinne Tuytelaars, and Andreas S
7 Pith papers cite this work, alongside 583 external citations. Polarity classification is still indexing.
7
Pith papers citing it
583
external citations · OpenAlex
citation-role summary
background 1
citation-polarity summary
years
2026 7verdicts
UNVERDICTED 7roles
background 1polarities
background 1representative citing papers
A cross-version swap protocol reveals dominant skills that swing composition success by up to 50 percentage points, and an atomic probe with selective revalidation governs updates at lower cost than always re-testing full compositions.
Entity interference from new knowledge graph entities causes up to 25% overestimation of CKGE method performance by disrupting prior predictions, requiring a corrected evaluation protocol.
Amnesia is a replay composition attack on continual learning that tilts class distributions under visibility (delta) and mass (f) budgets to reduce accuracy while evading audits.
BiCyc aligns old and new class representations bidirectionally with cycle consistency to preserve classification decisions and mitigate forgetting in exemplar-free continual learning.
MoLEM achieves a 10.40% average accuracy improvement in continual learning tasks across math, science, and code by using dynamic latent memory experts with a frozen base model and stage-specific autoencoders for routing.
DGMM is proposed as an explicit graph-structured memory architecture for AI that enables persistent episodic memory, cue-based recall, and context-dependent interpretation without retraining.
citing papers explorer
-
Continual Model Routing in Evolving Model Hubs
Formalizes continual model routing (CMR), releases CMRBench with over 2000 models, and presents CARvE which outperforms retrieval, fine-tuning and adapter-merging baselines on model/family/domain accuracy.
-
Atomic-Probe Governance for Skill Updates in Compositional Robot Policies
A cross-version swap protocol reveals dominant skills that swing composition success by up to 50 percentage points, and an atomic probe with selective revalidation governs updates at lower cost than always re-testing full compositions.
-
Revisiting Catastrophic Forgetting in Continual Knowledge Graph Embedding
Entity interference from new knowledge graph entities causes up to 25% overestimation of CKGE method performance by disrupting prior predictions, requiring a corrected evaluation protocol.
-
Amnesia: A Stealthy Replay Attack on Continual Learning Dreams
Amnesia is a replay composition attack on continual learning that tilts class distributions under visibility (delta) and mass (f) budgets to reduce accuracy while evading audits.
-
Two-Way Is Better Than One: Bidirectional Alignment with Cycle Consistency for Exemplar-Free Class-Incremental Learning
BiCyc aligns old and new class representations bidirectionally with cycle consistency to preserve classification decisions and mitigate forgetting in exemplar-free continual learning.
-
Dynamic Mixture of Latent Memories for Self-Evolving Agents
MoLEM achieves a 10.40% average accuracy improvement in continual learning tasks across math, science, and code by using dynamic latent memory experts with a frozen base model and stage-specific autoencoders for routing.
-
The Dynamic Gist-Based Memory Model (DGMM): A Memory-Centric Architecture for Artificial Intelligence
DGMM is proposed as an explicit graph-structured memory architecture for AI that enables persistent episodic memory, cue-based recall, and context-dependent interpretation without retraining.