EMem-Bench - Experience Generalization: leaderboard
Metric: Success rate (%; the 203 Experience Generalization episodes, applying a regularity learned from earlier corrections to a new object; 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 December 2026. 17 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash | 75.4 |
| 2 | Gemini 2.5 Pro | 72.9 |
| 3 | GPT-5.4 | 69 |
| 4 | Qwen 3.6 27B | 68 |
| 5 | Qwen 3 VL 32B Instruct | 63.1 |
| 6 | GPT-5.4 Mini | 41.4 |
| 7 | InternVL3-38B | 33 |
| 8 | Qwen 3 VL 8B Instruct | 21.7 |
| 9 | Qwen 3 VL 4B Instruct | 15.8 |
| 10 | InternVL3-8B | 4.4 |
Interactive version: theaggregate.ai/benchmark?slug=emem-bench-experience-generalization · How It Works · Data refreshed daily, snapshot 2026-09-26.