Memories That Outlive Their Reasons: Dependency-Tethered Memory for Tool-Using LLM Agents

Key points

• Agents now act on the Web, publishing packages, models and datasets through registry APIs, and remembering past verifications spares them the work.1
• When a signing key is revoked, everything built on it stops passing, yet the advisory names only the key, and the old verdicts of its dependents may never share a retrieval window with it.1
• On Fallout, a generator of versioned worlds with hypothesis-blind simulated histories shared by every memory, Tether publishes 0.3% of upstream-failing objects across three domains when relations are stated with their verdicts, with 37% fewer verification calls than checking everything.1
• These gains depend on capturing dependencies: when relations arrive in sessions apart from their verdicts, Tether's downstream unsafe publication rate rises to 9.4-22.1%, and a weak writer that misses edges loses safety in the same way.1

Sources (1)

• [1]Memories That Outlive Their Reasons: Dependency-Tethered Memory for Tool-Using LLM Agents
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 18, 09:01 AM

  This non-locality of invalidation turns memory into a liability: in long simulated chats, an agent reading a memory of atomic facts publishes 62% of the objects whose failure lies upstream, each backed by a record that was true when written.

  Agents now act on the Web, publishing packages, models and datasets through registry APIs, and remembering past verifications spares them the work.

Extractive summary: sentences quoted from the sources.