aiqbench analysis
This arXiv preprint proposes MemPursuit, a training-free method for selecting long-term agent memories by how much each adds beyond what has already been chosen, rather than by individual relevance alone. The reported gains (7.2 points at Top-10 on LoCoMo, and a 32.9x reranking latency speedup over a batched BGE reranker) are benchmark results from one paper and need independent replication across tasks and memory systems. For readers building agents, the practical question is whether this composition approach holds up against stronger retrievers and in production-scale memory stores.
Recall What Is Missing: Composing Evidence from Long-Term Agent Memory
arXiv:2610.15642v1 Announce Type: new Abstract: Under a limited context budget, a long-term agent must recall not merely memories that are individually relevant, but evidence that is collectively sufficient for the current query. Conventional memory access retrieves, scores, and truncates memories, often spending multiple context slots on redundant facts while leaving complementary evidence behind. We cast the read phase of long-term memory as query-conditioned evidence composition: each memory is valued by what it adds beyond the evidence already selected. We instantiate this operation with MemPursuit, a training-free mechanism that uses the residual of a sparse convex query reconstruction as a representation-space proxy for what the current memory set is missing. By reusing the query and candidate embeddings already available from retrieval, MemPursuit progressively composes nested evidence contexts without invoking a language model or cross-encoder to score each query--memory pair. The effectiveness of our method is verified on 3 popular benchmarks in comparison to strong baselines. In fixed-budget LoCoMo experiments, MemPursuit improves end-to-end accuracy by 7.2 percentage points at Top-10 and 5.5 percentage points at Top-20, with consistent gains across multiple memory frameworks, while achieving a 32.9$\times$ speedup in median reranking latency over fully batched BGE-v2-m3. Our implementation is available at https://github.com/FengQi-HITSZ/MemPursuit.
Abstract:Under a limited context budget, a long-term agent must recall not merely memories that are individually relevant, but evidence that is collectively sufficient for the current query. Conventional memory access retrieves, scores, and truncates memories, often spending multiple context slots on redundant facts while leaving complementary evidence behind. We cast the read phase of long-term memory as query-conditioned evidence composition: each memory is valued by what it adds beyond the evidence already selected. We instantiate this operation with MemPursuit, a training-free mechanism that uses the residual of a sparse convex query reconstruction as a representation-space proxy for what the current memory set is missing. By reusing the query and candidate embeddings already available from retrieval, MemPursuit progressively composes nested evidence contexts without invoking a language model or cross-encoder to score each query--memory pair. The effectiveness of our method is verified on 3 popular benchmarks in comparison to strong baselines. In fixed-budget LoCoMo experiments, MemPursuit improves end-to-end accuracy by 7.2 percentage points at Top-10 and 5.5 percentage points at Top-20, with consistent gains across multiple memory frameworks, while achieving a 32.9$\times$ speedup in median reranking latency over fully batched BGE-v2-m3. Our implementation is available at this https URL.
Artificial Intelligence (cs.AI)
arXiv:2610.15642 [cs.AI]
(or arXiv:2610.15642v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2610.15642 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 291 words
Related coverage
J++ Lens: Jacobian Filtering Enables More Faithful Workspace Lenses
LessWrong · Oct 8, 2026
Hint-based CoT faithfulness evals still mostly work on Claude
LessWrong · Jul 30, 2026
Jev: System One models for Prod, not God — with Diogo Almeida, CEO, TypeSafe AI
Latent Space · Sep 21, 2026
The Right Memory in the Wrong Context: Verifying Retrieval Admissibility in Long-Term Agent Memory
arXiv · Oct 7, 2026