# DISCERN: Preserve at Write, Resolve at Read for Persistent LLM-Agent Memory

DevFeed: [DISCERN: Preserve at Write, Resolve at Read for Persistent LLM-Agent Memory](<https://devfeed.tech/articles/discern-preserve-at-write-resolve-at-read-for-persistent-llm-agent-memory-132165.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.15729>)

Author: Chirag Ahuja, Vladislav Tanov, Ramcharan Kakarla, Jonathan Taws

Published: 2026-11-13T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Entity resolution](<https://devfeed.tech/topics/entity-resolution.md>), [Retrieval Augmented Generation (RAG)](<https://devfeed.tech/topics/retrieval-augmented-generation-rag.md>), [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [Enhancing AI model performance](<https://devfeed.tech/topics/enhancing-ai-model-performance.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [agent-memory](<https://devfeed.tech/tags/agent-memory.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-loop](<https://devfeed.tech/tags/agentic-loop.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [enterprise-agents](<https://devfeed.tech/tags/enterprise-agents.md>), [entity-resolution](<https://devfeed.tech/tags/entity-resolution.md>), [framework](<https://devfeed.tech/tags/framework.md>), [graph](<https://devfeed.tech/tags/graph.md>), [llm](<https://devfeed.tech/tags/llm.md>), [llm-as-a-judge](<https://devfeed.tech/tags/llm-as-a-judge.md>), [memory](<https://devfeed.tech/tags/memory.md>), [navigation](<https://devfeed.tech/tags/navigation.md>), [rag](<https://devfeed.tech/tags/rag.md>)

## AI overview

DISCERN targets persistent-agent memory resolution rather than retrieval: identifying the intended entity, the current time-state, and when to abstain. It preserves information in an additive bi-temporal provenance graph, keeping confusable entities separate and retaining superseded values; at read time, an agent navigates the graph through a zero-LLM tool surface and chooses among four gated terminal actions. On the 310-query synthetic ENT-310 benchmark, it scores 0.924 abstention F1 versus at most 0.802 for eight baselines and 0.861 macro answer correctness versus 0.806 for the strongest baseline, with the largest gain in entity resolution. With no tuning on the full LoCoMo corpus, it also leads macro correctness, 0.663 versus 0.638, and remains among the best calibrated on false-premise abstention. These results are reported in the supplied abstract.

## Source excerpt

arXiv:2611.15729v1 Announce Type: new Abstract: Enterprise agents accumulate memory across dozens of sessions in which facts contradict, get superseded, and blur as similarly named entities collide in embedding space. The field treats this as a retrieval problem; we show it is not. On ENT-310, a synthetic enterprise sales-agent memory benchmark of 310 queries (210 answerable, 100 unanswerable), a vanilla Retrieval-Augmented Generation (RAG) reader over a memory store already matches strong memory systems on the recall pillars. Retention and currency are near-saturated, roughly 0.93 to 0.98 for every competent system, so recall is essentially solved. What remains unsolved is resolution: which entity a query refers to, which time-state is current, and whether to commit an answer at all or decline when evidence is insufficient. We propose DISCERN (DISambiguation and Cross-session Episodic memoRy Navigation), an agent-memory framework built on a single rule: preserve at write, resolve at read. At write time DISCERN builds a strictly additive bi-temporal provenance graph that never merges confusable entities and never drops superseded values. At read time a zero-LLM tool surface drives an agentic loop that navigates this graph and commits exactly one of four gated terminal actions. A four-judge LLM-as-a-Judge council scores two axes: answer correctness on answerable questions, and the answer-versus-abstain decision as a classification problem. On ENT-310, DISCERN is the best-calibrated system on abstention (F1 0.924 versus at most 0.802 across eight baselines) and leads macro answer correctness under equal pillar weighting (0.861 versus 0.806 for the strongest baseline), a lead driven by entity resolution (0.800, where no baseline exceeds 0.40). Applied with zero tuning to the full LoCoMo corpus, the same pipeline again leads macro answer correctness (0.663 versus 0.638) and stays among the best-calibrated on false-premise abstention.