Spaces:
Running
Running
| title: AgentDeck Agentic Edge Viewer | |
| emoji: 🎮 | |
| colorFrom: blue | |
| colorTo: green | |
| sdk: static | |
| app_file: index.html | |
| pinned: false | |
| license: mit | |
| short_description: Curated AgentDeck study match replays. | |
| # AgentDeck Agentic Edge Viewer | |
| Interactive replay viewer for five curated matches from **The Agentic Edge: Strategy Stack Effects on LLM Agency in Sequential Decision Environments**. | |
| Use this Space to see how the study's behavioral evidence looks inside recorded matches: | |
| 1. **S0 failure** — FlashLite attacks until it dies with all potions unused. | |
| 2. **S1 pivot** — structured reasoning changes the same threshold moment into a healing decision. | |
| 3. **S3 policy execution** — HP grounding produces traceable policy behavior. | |
| 4. **VariableDamage risk policy** — risk-band grounding transfers to stochastic damage. | |
| 5. **VariableDamage caveat** — a correct-looking policy still loses when seat/game dynamics dominate. | |
| ## Study Artifacts | |
| - Dataset: https://huggingface.co/datasets/agentdeck/agentic-edge-strategy-stack-study | |
| - GitHub: https://github.com/agentdeck/agentdeck-core | |
| The dataset is the canonical artifact store. This Space is the visual companion for exploring selected matches. | |