AppWorld

The 'AppWorld' dataset is a benchmark that contains a collection of agentic traces used to evaluate the performance and scalability of prompt learning methods for language model agents.

Papers using AppWorld (12)

Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models2025 · 2 citesToward Scalable Verifiable Reward: Proxy State-based Evaluation For Multi-turn Tool-calling LLM Agents2026 · 2 citesNot All Skills Help: Measuring and Repairing Agent Knowledge2026ACCORD: Action-Conditioned Contextual Grounding for Language Agents2026Hera: Learning Long-Horizon Coordination for Device-Cloud Collaborative LLM Agents2026Agents Explore but Agents Ignore: LLMs Lack Environmental Curiosity2026SeeUPO: Sequence-Level Agentic-RL with Convergence Guarantees2026Combee: Scaling Prompt Learning for Self-Improving Language Model Agents2026SkillX: Automatically Constructing Skill Knowledge Bases for Agents2026ACON: Optimizing Context Compression for Long-horizon LLM Agents2025Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models2025Reinforcement Learning for Self-Improving Agent with Skill Library2025