Georgia Tech · Embodied AI

Georgia Tech benchmarks robot memory across day-long tasks

Benchmark studySeptember 24, 2026

Concept illustration of a cutaway virtual home with several rooms, a small mobile robot, translucent earlier object positions and a subtle sequence of remembered locations; clearly a simulated environment, no false benchmark screenshot — **AI-generated illustration** — Conceptual illustration of robot memory in a simulated home; not a photograph of the reported prototype or experiment. Created with OpenAI image generation · 1536 × 1024 · Illustrative concept, not a photograph of the reported event.

What happened

Georgia Tech researchers evaluated models including Gemini 2.0 Flash, GPT-4o and an open-source Qwen reasoning model. The university reports that Qwen performed best on the high-difficulty tasks; the work was presented at the 2026 European Conference on Computer Vision.

Why it matters

Long-term visual memory is important for robots that must remember where objects moved and what happened earlier. This is a benchmark evaluation, not a physical household deployment or proof of reliable long-term operation.

Source checked October 31, 2026. Status describes the evidence in the linked source at the time of review.