EmbodiedCLBench: Evaluating Continual Learning for Self-Evolving Embodied Agents
To address this gap, we introduce EmbodiedCLBench, the first benchmark in embodied simulation for evaluating the continual learning capability of self-evolving agents.
ProofPaper ↗
Key points
- Continual learning requires models to accumulate past experience and transfer it to new tasks.
- With compositional generalization as the core evaluation principle, our benchmark evaluates whether agents can first learn from basic tasks and then recombine the acquired skills to solve complex, unseen advanced tasks.
- Based on EmbodiedCLBench, we conduct extensive experiments across multiple agent harnesses and learning methods, revealing consistent limitations and distinctive behaviors.
- Collectively, our findings highlight continual learning as a fundamental yet underexplored capability, and our benchmark offers valuable resources for designing effective self-evolving agents in embodied environments.
Sources (2)
- [1]EmbodiedCLBench: Evaluating Continual Learning for Self-Evolving Embodied AgentsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 9, 05:23 AM
To address this gap, we introduce EmbodiedCLBench, the first benchmark in embodied simulation for evaluating the continual learning capability of self-evolving agents.
Continual learning requires models to accumulate past experience and transfer it to new tasks.
- [2]EmbodiedCLBench: Evaluating Continual Learning for Self-Evolving Embodied AgentsHugging Face Daily Papers · Oct 9, 12:00 AM · same content
Extractive summary: sentences quoted from the sources.
Before this
- Oct 8, 2026REDaCT: Collapse-Free Unlearning in Large Reasoning Models
- Oct 8, 2026openai/codex rust-v0.162.0: 0.162.0
- Oct 8, 2026Show HN: Jevman – AI decision models play Pac-Man
- Oct 8, 2026Microsoft Joins the Local AI Push
- Oct 8, 2026Understanding Jev, the new model everyone is talking about
- Oct 7, 2026[AINews] Claude Haiku 5.5 — better than GPT-6 Luna at the same pricing