The report audits a long‑term‑memory retrieval chain on 500 LongMemEval‑S development questions, evaluating reader performance under an adapted GPT‑4o rubric. Scores vary widely across reader lanes (93–479), with the strongest lanes scoring 479 and 475, and re‑judging the same pass‑1 answers changes three labels, yielding 478. Fixed‑answer knowledge‑update re‑scoring achieves 70/72 or 69/72 depending on the template, and a different‑family reader scores 474, just 1.0 percentage point below the headline pass. The study also reports that live reader request bodies were not retained, that 18 of 23 gains occurred where baseline packets lacked evidence, and that a negative control rejects a verifier that repairs some wrong drafts but breaks many correct ones. Overall, the findings do not establish a new leaderboard leader or a transferable memory advantage, and the released artifacts support packet inspection and re‑scoring but do not reconstruct the method.
arXiv:2609.38021v1 Announce Type: cross Abstract: We evaluate an auditable long-term memory system on LongMemEval-S. Its retrieval chain uses hybrid candidate retrieval, cross-encoder reranking, cove...
The paper reports a preregistered audit of language‑model judges used as measurement instruments, revealing that the assumption that a model’s responses remain stable over time is invalid. Across nearly 53,000 audited requests, repeat rankings and byte‑identical replays fell far below required reliability thresholds, with three identified mechanisms—label‑to‑meaning bias, candidate gaps below the noise floor, and input permutation noise—explaining the discrepancy. The study proposes a three‑level snapshot‑identity framework, eight design rules, and a reporting checklist to prevent such reliability failures in future evaluations.
The paper investigates how memory systems can answer a current query correctly yet fail to retain distinctions needed for later updates. Using a paired‑history audit, the authors evaluate 24 history pairs across six synthetic mechanisms and two model backends, achieving perfect reveal accuracy on DeepSeek and high accuracy on GLM. Record‑level audits reveal specific failures in structured reveal memories and frontier late‑reference adequacy, and the authors test a label‑equivariant repair that only partially restores correctness.
The paper introduces LSREP, a Longitudinal State‑Replay Evaluation Protocol designed to assess how conversational memory evolves over time, incorporating ordered replay, lifecycle schedules, repeated probes, evolving reference answers, and mechanism‑fidelity checks. It applies LSREP to ICE v2, a local‑first memory middleware, and reports that on three ordinary‑density datasets ICE v2 achieves near‑zero mean quality difference from vector‑RAG while using fewer fragments but slightly more prompt tokens, yet fails catastrophically on a dense dataset. In a public diagnostic, ICE v2 underperforms pure vector‑RAG on LongMemEval, revealing significant multi‑session and temporal failures and a quality‑cost trade‑off rather than superior efficiency.
arXiv:2609.08279v1 Announce Type: cross Abstract: Agent memory systems must discard stored information when their history exceeds a fixed token budget. Existing budget-accuracy frontiers quantify the...
The paper presents a retrieval‑admissibility verification framework for long‑term memory agents, classifying each memory‑query pair as admissible, inadmissible, or unresolved. It evaluates the framework on public benchmarks (RHELM and MemOps), showing improved anchor recall and reduced exact similarity errors, while also revealing that existing verifiers miss certain inadmissible exposures. The study highlights the need for separate checks on candidate support, admissibility, prompt exposure, and answer disclosure to ensure safe memory retrieval.
arXiv:2606. 09900v1 Announce Type: cross Abstract: Long-term memory is the missing layer for LLM agents: across sessions they forget, and the common workaround -- replaying the whole history into the prompt -- is expensive, slow, and, as distractors accumulate, less accurate.
The paper investigates the reliability of language‑model judges used as measurement instruments on shared endpoints. Through two preregistered audits of 52,988 requests, the authors found that repeat rankings and byte‑identical replays fell far short of required thresholds, revealing significant instability. They identify three mechanisms—label‑to‑meaning bias, candidate gaps below the noise floor, and input permutation noise—that explain the gap, and propose a snapshot‑identity ladder, design rules, and a reporting checklist to mitigate such failures.
The paper investigates whether the object selected in a grounded language‑model pipeline actually reaches the reader, a failure that can break the handoff between stages. By auditing 600 HybridQA questions across three selector families, the authors find that exact key lookup and title matching recover the selected object in all 1,463 resolvable records, but body‑only BM25 omits it in 26.6% of cases at cutoff five, while hybrid retrieval with reranking omits it only 1.0%. The study also shows that misalignment between selected and retrieved objects can reduce exact match scores by up to 31 points, and introduces the Returned‑Object Profile (ROP) as a tool for reproducible auditing.
arXiv:2608. 08944v1 Announce Type: cross Abstract: A failed retrieval-augmented generation (RAG) answer can be consistent with several unseen responses to evidence repair.
The paper investigates how language‑model judges can make version‑dependent errors when evaluating upgraded agents. Using 35 public coding‑agent submissions, two customer‑service agents, and over a thousand expert‑labeled trajectories, the authors show that fixed judges often reject task‑conditioned error invariance and can incorrectly approve failed patches, especially as agent capability increases. Paired audits of current outputs reduce interval width only marginally, and the study concludes that independent human patch review is still necessary.
The paper reports on a series of experiments examining how different forms of directives—such as record pointers, criteria, or combinations—affect an agent’s choice of archived source records when it inherits six one-line memories. Across twelve registered studies involving 14,760 attempts on a single instrument lineage, the authors measured the impact of various directive formats on six direct-provider models, nine OpenRouter-served models, and several Claude and Opus 5 models, noting differences in performance metrics and replication outcomes. The results are purely descriptive, detailing the effects of exact edits on fixed panels with registered intervals and no claim of underlying mechanisms.