aiqbench analysis
ACCT proposes reusing a context's key-value (KV) cache across LLMs of different sizes within one model family, using a sparse transformation so the target model can skip reprocessing the original context. The abstract reports matched or better performance with a 75% smaller KV cache and a 6.5-21.2x speedup over target-model prefill at about 19K tokens, but these figures come from the authors' own benchmarks and need independent replication. For practitioners, the approach matters mainly for heterogeneous serving pipelines that route the same long context to models of different scales.
ACCT: Asymmetric Cross-LLM Context Transfer
arXiv:2610.42116v1 Announce Type: new Abstract: In real-world deployments, large language model (LLM) systems are increasingly evolving from single-model invocation toward heterogeneous deployments in which different models are dynamically invoked for inference. When the same context is reused across models, however, each newly selected target model typically has to process it again, repeatedly incurring substantial computational costs. Therefore, we propose ACCT (Asymmetric Cross-LLM Context Transfer), which targets KV cache transfer between models of different scales within the same model family. ACCT transforms the KV cache produced by a source model into sparse KV states that can be directly consumed by a target model, thereby enabling cross-model reuse of context computation and decoupling context computation from downstream inference. ACCT consists of three lightweight components: the Cartographer identifies query-relevant context tokens for each target layer and KV head; the Allocator determines how many context tokens each target layer and KV head retains under a shared total KV budget; and the Compiler transforms selected source KV into sparse target KV. The target model can therefore perform inference using the transferred KV without reprocessing the original context. We evaluate ACCT in two complementary settings comprising five general-capability benchmarks and six highly context-dependent tasks. Experiments show that ACCT matches or exceeds the target model's original inference performance even when reducing the KV cache size by 75%. At a context length of approximately 19K tokens, ACCT's KV conversion achieves a 6.5-21.2 times speedup over target context prefill. These results demonstrate the feasibility of transferring context computation across models under a limited KV budget.
Abstract:In real-world deployments, large language model (LLM) systems are increasingly evolving from single-model invocation toward heterogeneous deployments in which different models are dynamically invoked for inference. When the same context is reused across models, however, each newly selected target model typically has to process it again, repeatedly incurring substantial computational costs. Therefore, we propose ACCT (Asymmetric Cross-LLM Context Transfer), which targets KV cache transfer between models of different scales within the same model family. ACCT transforms the KV cache produced by a source model into sparse KV states that can be directly consumed by a target model, thereby enabling cross-model reuse of context computation and decoupling context computation from downstream inference. ACCT consists of three lightweight components: the Cartographer identifies query-relevant context tokens for each target layer and KV head; the Allocator determines how many context tokens each target layer and KV head retains under a shared total KV budget; and the Compiler transforms selected source KV into sparse target KV. The target model can therefore perform inference using the transferred KV without reprocessing the original context. We evaluate ACCT in two complementary settings comprising five general-capability benchmarks and six highly context-dependent tasks. Experiments show that ACCT matches or exceeds the target model's original inference performance even when reducing the KV cache size by 75%. At a context length of approximately 19K tokens, ACCT's KV conversion achieves a 6.5-21.2 times speedup over target context prefill. These results demonstrate the feasibility of transferring context computation across models under a limited KV budget.
Artificial Intelligence (cs.AI)
arXiv:2610.42116 [cs.AI]
(or arXiv:2610.42116v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2610.42116 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 · 345 words