Agents can retain old facts without using them indiscriminately

RD-Forget preserves an agent’s full observation history while constructing a query-specific memory view that suppresses obsolete facts when appropriate.

PaperTop Universitycs.AIarXiv:2609.10263v1
Yuhang Li · Yuchen Li

Beihang University · East China Normal University

Research Digest··2 min read
Li and Li introduce RD-Forget, a training-free memory framework that separates long-term storage from the evidence used to answer each query. Across five memory and reasoning settings, the authors report that query conditioning and selective suppression of superseded facts improve answers while preserving access to historical information.

What they did

The authors retain all observations in a source archive, then use a frozen language model to build a smaller, query-conditioned memory view at answer time. The curator extracts relevant evidence, groups facts into semantic slots, links newer values to the same-slot values they replace, and retains relations needed for multi-step reasoning.

A rate-distortion objective—which balances memory size against loss of useful information—guides evidence selection under a fixed memory budget. The shared answering pipeline was evaluated on conversational memory, knowledge updating, fact consolidation, long-context reasoning, and personalization tasks.

Key findings

  • Configurations lacking selective forgetting or query conditioning produced the largest score deficits in the reported ablations.
  • Replacement links helped prevent superseded values from influencing questions about the current state.
  • Intent-aware retrieval restored older evidence for historical questions rather than deleting it permanently.
  • Semantic-slot grouping and relation preservation made complementary contributions, including support for multi-hop reasoning.

Why it matters

The work reframes forgetting as control over information use rather than deletion from storage. This distinction could help persistent agents handle changing user preferences and world facts without sacrificing historical recall or contaminating current answers with obsolete alternatives.

Caveats

RD-Forget depends on a frozen language model to identify evidence, semantic slots, replacement relations, and query intent, so curator errors could propagate to the final answer. The abstract reports broad evaluations and relative ablation trends but provides no numerical effect sizes, leaving efficiency, robustness across model families, and performance under ambiguous or contradictory updates as open questions.

§

Analysis

The paper addresses a central tension in agent memory: immutable archives support auditability and historical questions, while unfiltered retrieval can expose the model to stale or conflicting evidence. RD-Forget’s layered design resembles database-style separation between stored records and query-time views, adapted to language-model memory.

A key next question is whether the framework remains reliable when facts do not fit cleanly into slots, updates are only partially superseding, or user intent is ambiguous. Measuring curator cost and failure modes will also be important for long-running agents with rapidly growing archives.

newspaper

Research Digest

Articles published under the Zotpaper byline are synthesized from multiple source publications by our AI editor and reviewed by our editorial process. Each story combines reporting from credible outlets to give readers a balanced, comprehensive view.