CAS opens ZDTaichu5.0-9B: 9B multimodal model leads spatial benchmarks by 20 points
新智元 · wechat · 2026-09-16
ZDTaichu 5.0-9B, an open-source multimodal model from CAS, tops 8 of 9 spatial benchmarks in its size class (MindCube-tiny 78.27, +20.67 pts over Qwen3.5-9B) without losing general skills, using entropy-gated recurrent inference and a progressive data curriculum — weights and training recipes are open, with real-robot lab and factory demos.
More from Embodied
- Drone caught on camera giving a dog a bath in China — ZeroStateReflex · 2026-09-16
- RoboDojo Puts GPT-6 Astra to the Test: Strong Semantics, Weak Physical Commonsense — ZeYanjie · 2026-09-16
- Muse Now Picks Its Own Outfit Every Morning, Scale CEO Approves — alexandr_wang · 2026-09-16
- Scoble predicts billions of evolving personas per person, ahead of Meta Connect avatar demo — Scobleizer · 2026-09-16
- Hacked computer-use for iPad: USB dongle bypasses iPadOS accessibility limits — alfcnz · 2026-09-16
- Apple's Reference Image hashes raw negatives in-sensor for verifiable photography — nptacek · 2026-09-16