Dev teaches Qwen2.5-VL-3B to drive in TrackMania from 300 replays, no RL needed
k7agar · x · 2026-10-23
A developer trained Qwen2.5-VL-3B to play TrackMania (AO1): the vision-language model watches frames, reads car telemetry and presses keys. No reinforcement learning — it learned purely from 300 expert replays via behavior cloning. It finishes about half its laps with a best lap of 24.6s.
More from Fun
- The Anthropic Applicant Who Needs 20 Years in the Woods First — kellerjordan0 · 2026-10-23
- 3 devs spent 2 years building an AI-assisted roguelike horror card game, demo on Steam — AlvinhoGames_ · 2026-10-23
- 'Woken by the token reset': developers joke about living on AI usage quotas — burhop · 2026-10-23
- AI-generated movie poster 'The Chase' crams a wolf, a plane and muscle cars into one chase — Aggravating-Sir-115 · 2026-10-23
- Reporter Loops an AI Agent to Hoard Shopping Cart Holds for Hours — lfschiavo · 2026-10-23
- Dating app rejection comes back with only an 'InvalidDateException' error message — gregmushen · 2026-10-23