analiza aiqbench
This research establishes theoretical bounds for pre-query retention in agents, optimizing how information is encoded and retrieved from immutable archives under strict bit or word read constraints. By characterizing feasible memory cells and comparing explicit encoders against decision-tree counting methods, the study provides rigorous limits on storage efficiency and decoding latency for future dialogue systems. The work bridges information theory and AI architecture to define the fundamental trade-offs between retention capacity and retrieval bandwidth.
Pre-Query Retention with Bounded Bit and Word Retrieval
arXiv:2610.25098v1 Announce Type: cross Abstract: An agent can retain information before receiving a future question and later retrieve evidence from an archive. Evaluating this arrangement requires accounting for both retained information and the width of each retrieval response. Decision-tree capacity bounds describe individual sets of distinguishable records, but a memory writer must encode every possible record before the query is known. We study fixed-length retention beside an immutable binary archive, with a worst-case limit on raw bit or word reads. We characterize feasible one-bit-read memory cells and construct encoders that partition the complete record space. Supporting pairwise exclusive disjunction and conjunction with one bit read is equivalent to supporting every Boolean query. For multiple reads, we combine explicit encoders with decision-tree counting and an entropy converse. The word-read bounds differ by at most two retained bits when the read budget is at most half the archive length. Under an explicit condition on word width and archive length, raw-word caching attains the exact optimum among all writers. A related condition preserves this optimum under positive average error. Exhaustive checks validate the implementations. In a matched dialogue-agent study, a categorical sufficient statistic preserves source decisions with less storage and lower decoding latency than the universal encoder. A catalogue-specific index reduces private state further and requires additional shared storage. The results characterize how response width and future query structure determine retention requirements.
Abstract:An agent can retain information before receiving a future question and later retrieve evidence from an archive. Evaluating this arrangement requires accounting for both retained information and the width of each retrieval response. Decision-tree capacity bounds describe individual sets of distinguishable records, but a memory writer must encode every possible record before the query is known. We study fixed-length retention beside an immutable binary archive, with a worst-case limit on raw bit or word reads. We characterize feasible one-bit-read memory cells and construct encoders that partition the complete record space. Supporting pairwise exclusive disjunction and conjunction with one bit read is equivalent to supporting every Boolean query. For multiple reads, we combine explicit encoders with decision-tree counting and an entropy converse. The word-read bounds differ by at most two retained bits when the read budget is at most half the archive length. Under an explicit condition on word width and archive length, raw-word caching attains the exact optimum among all writers. A related condition preserves this optimum under positive average error. Exhaustive checks validate the implementations. In a matched dialogue-agent study, a categorical sufficient statistic preserves source decisions with less storage and lower decoding latency than the universal encoder. A catalogue-specific index reduces private state further and requires additional shared storage. The results characterize how response width and future query structure determine retention requirements.
24 pages, 4 figures. Complete proofs and experimental details included. Both authors contributed equally
Information Theory (cs.IT); Artificial Intelligence (cs.AI)
arXiv:2610.25098 [cs.IT]
(or arXiv:2610.25098v1 [cs.IT] for this version)
https://doi.org/10.48550/arXiv.2610.25098 Focus to learn more arXiv-issued DOI via DataCite
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.
Text integral sub licența deschisă a publicației · 333 cuvinte
Subiecte conexe
When Retrieval Fails Before It Begins: Structurally Indirect Prerequisite Eviction as a Retention Failure in Agentic Memory
arXiv · 24 aug. 2026
What Should an Agent Remember? Disentangling Retention from Retrieval in Bounded-Memory Evaluation
arXiv · 2 oct. 2026
Learning What to Remember: Observability-Safe Memory Retention via Constrained Optimization for Long-Horizon Language Agents
arXiv · 17 iun. 2026
Kernel-Based Steering of CLIP with Vision-Language Model Preferences
arXiv · 30 sept. 2026