caseclose

caseclose/cma-harness

Cognitive-structured Multimodal Agent (CMA-Harness): a memory-centric agent for long-horizon multimodal understanding, generation, and editing — externalizing visual history into episodic memory with selective retrieval. Includes the M2CA-Bench benchmark.

⭐ 14 ⑂ 2 CSS · 4 天前推送
14
Watchers
0
贡献者
0
Commits
0
Releases
1
Open Issues
4 天前
最近推送
原文 中文
暂无 README