WorldMemBench: 260-case benchmark tests video world models' memory across revisits
ZJU-REAL · hf · 2026-10-16
Researchers from Zhejiang University, Fudan, Ant Group and Nanjing University released WorldMemBench, a benchmark for evaluating world memory in video world models, on arXiv.
- 260 cases across 23 categories covering human, object, environment and dynamic dimensions.
- Uses a multi-revisit paradigm to test static memory (identity, appearance, layout, depth) and dynamic retention across state transitions.
- Provides capability profiles of nine models, exposing weaknesses in consistency across repeated revisits.
More from Research
- Sakana AI scientist to keynote on self-improving AI at Seoul's DEEP DIVE INTO 2026 — Stefania_druga · 2026-10-16
- Lean v4.35.0 ships verified kernels — one written and proven consistent by Claude — geoffreyirving · 2026-10-16
- MP-Bench: 12 voice agents no better than coin flip at knowing when to speak in group chats — solyarisoftware · 2026-10-16
- Neuron-level GWAS + PheWAS on Qwen3-8B pinpoints safety-critical neurons — Sauers_ · 2026-10-16
- Tencent squeezes 770B-parameter Hunyuan Hy4 Preview into 214 GiB at ~2.38 bits per weight — coolmum18 · 2026-10-16
- 200 chess games suggest in-context learning doesn't work yet, experimenter says — akbirthko · 2026-10-16