HiVe-Mem: NTU-led framework for multimodal agent long-term memory
NanyangTechnologicalUniversity · hf · 2026-11-13
Researchers from Nanyang Technological University, Tsinghua, and Sun Yat-sen University introduce HiVe-Mem, a memory framework for multimodal agents handling long interaction histories.
- Existing systems rely on semantically compressed memories, missing relevant info and heterogeneous evidence needs
- Horizontal Memory Expansion: organizes memory episodes into a relational graph with textual/visual anchors, expanding retrieved seeds with related neighbors
- Vertical Evidence Composition: each memory retains multiple evidence forms (summaries, captions, grounded visual views) composed adaptively per query
- On arXiv since Nov 9, 2026
More from coding & agent
- OpenAI Codex CLI 0.169.0 ships multi-tab side chats, task search, and custom provider onboarding — github-actions[bot] · 2026-11-13
- kagent 1.0 GA: Agent Substrate brings sandboxed AI agent execution to Kubernetes beyond pod-per-agent — bibryam · 2026-11-13
- Three quotumorg Hermes plugins land in NousResearch marketplace: precedent rulings and exfil blocking — Teknium · 2026-11-13
- Matt Pocock wants every important PR to ship with a demo video, building an /animatic skill — mattpocockuk · 2026-11-13
- ArtCraft clones Photoshop with AI as open source, has shipped 17 tools and wants to 'rebuild everything' — altryne · 2026-11-13
- Solo dev ships logic-puzzle roguelike built with Claude that screenshot-tests itself — TheCodingTutor · 2026-11-13