arXiv Artificial Intelligence

Consolidator: Learning Persistent Routed Memory Across Context Boundaries

1 hour ago 13.08.2026 07:00 👁 2 views
Consolidator: Learning Persistent Routed Memory Across Context Boundaries

Quick summary

arXiv:2608.11701v1 Announce Type: cross Abstract: Copying short-term memory (STM) into a slower store can preserve state across a context boundary, but persistence alone does not ensure that the retained state influences subsequent memory access. We test this distinction in a Phasor Memory Network (PMNet) using Consolidator, a shared slot-local operator that transforms routed STM before accumulating it into long-term memory (LTM), without replaying the source tokens. After each consolidation, the KV cache and STM are cleared. The retained LTM can still be read and is also fed into the hierarch

Key takeaways

  • arXiv:2608.11701v1 Announce Type: cross Abstract: Copying short-term memory (STM) into a slower store can preserve state across a context boundary, but persistence alone does not ensure that the retained state influences subsequent memory access.
  • We test this distinction in a Phasor Memory Network (PMNet) using Consolidator, a shared slot-local operator that transforms routed STM before accumulating it into long-term memory (LTM), without replaying the source tokens.
  • After each consolidation, the KV cache and STM are cleared.

Why it matters

The importance of “Consolidator: Learning Persistent Routed Memory Across Context Boundaries” will be measured by what changes in practice. User behavior, access conditions, verifiable performance and responsible-use outcomes are the signals worth following.

Kaynak sitede devamını oku: arXiv Artificial Intelligence ↗