| name | kage-bench-fast-known-axis-visual-generalization |
| title | KAGE-Bench: Fast Known-Axis Visual Generalization Evaluation for Reinforcement Learning Agents |
| version | 0.0.2 |
| engine | skillxiv-v0.0.2-claude-opus-4.6 |
| license | MIT |
| url | https://arxiv.org/abs/2601.14232 |
| keywords | ["Agent","Learning"] |
| description | Pixel-based reinforcement learning agents often fail under purely visual distribution shift even when latent dynamics and rewards are unchanged, but existing benchmarks entangle multiple sources of shift and hinder systematic analysis. We introduce KAGE-Env, a JAX-native 2D platformer that factorizes the observation process into independently controllable visual axes while keeping the underlying control problem fixed. By construction, varying a visual axis affects performance only through the in... |
Problem
KAGE-Bench addresses key challenges in autonomous agent development. This paper provides solutions for evaluating, building, or improving agent systems.
Key Approach
The paper introduces a novel framework, methodology, or benchmark for kage-bench. The core contributions include:
- Systematic framework or benchmark for agent evaluation and development
- Empirical findings on agent performance, efficiency, or capabilities
- Generalizable principles applicable across domains
When to Use
Use this skill when you need to:
- Evaluate or benchmark autonomous agent systems
- Understand best practices in agent design and evaluation
- Learn empirical results on agent performance
- Improve agent efficiency, reasoning, or capabilities
When NOT to Use
- For non-agent-related tasks
- When seeking quick implementation code (see the paper for details)
- For general knowledge unrelated to autonomous agents
Resources
See the paper for comprehensive methodology, experimental protocols, benchmarks, and implementation details.
