EMem-Bench - Dynamic Tracking: leaderboard
Metric: Success rate (%; the 1052 Dynamic Tracking episodes, acting on the latest location or state of an object that moved or changed; full-context condition: the complete interaction history as sanitized text plus the current 500 x 500 RGB view, one primitive action per call with deterministic decoding and at most eight target-task steps in AI2-THOR and ProcTHOR homes; an episode succeeds only if the simulator reaches the target terminal state). Source: arxiv.org. Saturation forecast: Around October 2027. 17 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash | 53.9 |
| 2 | Gemini 2.5 Pro | 51.6 |
| 3 | Qwen 3.6 27B | 49.5 |
| 4 | GPT-5.4 | 46.7 |
| 5 | Qwen 3 VL 4B Instruct | 37.3 |
| 6 | InternVL3-8B | 31.8 |
| 7 | Qwen 3 VL 8B Instruct | 26.6 |
| 8 | GPT-5.4 Mini | 25.8 |
| 9 | Qwen 3 VL 32B Instruct | 20.3 |
| 10 | InternVL3-38B | 17 |
Interactive version: theaggregate.ai/benchmark?slug=emem-bench-dynamic-tracking · How It Works · Data refreshed daily, snapshot 2026-09-26.