AdithyaSK/beyond-benchmarks-slides
0
Beyond Benchmarks: Building RL Environments for AI Agents
How raw data becomes an RL environment an agent can be trained in: a GeoGuessr game from street-level panoramas, a dataset of 5,394 data-analysis tasks, the same tasks inside Claude Code and Codex, Kannada OCR and ASR, and a container quay rebuilt from real Port of Barcelona records.
Download: PDF · PPTX, with video
Built with research-presentation-template.
Keys: ←/→ navigate · t dark/light · f fullscreen · p print to PDF · ⚙ slide list.
