analiza aiqbench
Attuner addresses computational inefficiency in LLM agents by enabling high-fidelity reuse of cached key-value states without recomputation or full-context prefill. The method achieves this through query-side adaptation using low-rank adapters, recovering performance lost by position-independent caching while training on less than 0.05% of model parameters. This approach significantly accelerates inference for tasks involving repeated loading of skills, documents, and memory entries.
ATTUNER: Recomputation-Free KV Cache Reuse via Query-Side Adaptation
arXiv:2609.36722v1 Announce Type: new Abstract: Large language model (LLM) agents repeatedly load reusable content, such as skills, documents, and memory entries, into the current context. Re-encoding this content for every request wastes computation. Position-independent caching (PIC) alleviates this by encoding each artifact independently and reusing its key-value (KV) states at arbitrary positions, but it incurs a quality loss relative to full-context prefill. Existing methods repair this loss by restoring global position IDs or recomputing selected tokens. In this work, we isolate the source of the loss, finding that the positional mismatch has minor effect, and independently cached artifacts retain faithful representations: reading a provided artifact stays largely accurate, and performance degrades only when the model must select among multiple artifacts. Moreover, replacing PIC's attention scores with full-prefill scores recovers performance with the cached KV unchanged, localizing the failure to the attention rather than KV recomputation. Motivated by this, we propose \textsc{Attuner}, a query-side adaptation method that learns to read a frozen artifact cache. \textsc{Attuner} inserts low-rank adapters into the query projections and is trained by distilling full-prefill distribution into the student. It trains fewer than 0.05\% of the model parameters and, at inference, requires neither cache recomputation nor a full-context reference. On Qwen3-4B and Qwen3-8B across seven benchmarks covering skills, documents, memory, and code, \textsc{Attuner} substantially outperforms prior PIC baselines in both in-domain and out-of-domain settings, matches full-context prefill quality while providing up to $3.73\times$ speedup.
Abstract:Large language model (LLM) agents repeatedly load reusable content, such as skills, documents, and memory entries, into the current context. Re-encoding this content for every request wastes computation. Position-independent caching (PIC) alleviates this by encoding each artifact independently and reusing its key-value (KV) states at arbitrary positions, but it incurs a quality loss relative to full-context prefill. Existing methods repair this loss by restoring global position IDs or recomputing selected tokens. In this work, we isolate the source of the loss, finding that the positional mismatch has minor effect, and independently cached artifacts retain faithful representations: reading a provided artifact stays largely accurate, and performance degrades only when the model must select among multiple artifacts. Moreover, replacing PIC's attention scores with full-prefill scores recovers performance with the cached KV unchanged, localizing the failure to the attention rather than KV recomputation. Motivated by this, we propose \textsc{Attuner}, a query-side adaptation method that learns to read a frozen artifact cache. \textsc{Attuner} inserts low-rank adapters into the query projections and is trained by distilling full-prefill distribution into the student. It trains fewer than 0.05\% of the model parameters and, at inference, requires neither cache recomputation nor a full-context reference. On Qwen3-4B and Qwen3-8B across seven benchmarks covering skills, documents, memory, and code, \textsc{Attuner} substantially outperforms prior PIC baselines in both in-domain and out-of-domain settings, matches full-context prefill quality while providing up to $3.73\times$ speedup.
Computation and Language (cs.CL)
arXiv:2609.36722 [cs.CL]
(or arXiv:2609.36722v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2609.36722 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.
Text integral sub licența deschisă a publicației · 326 cuvinte
Subiecte conexe
Optimizing ATOM and vLLM-ATOM for High-Interactivity Inference
AMD · 1 sept. 2026
PySpark Optimization: 12 Proven Techniques to Speed Up Your Spark Jobs
Analytics Vidhya · 27 mai 2026
Introducing ROCm™ AMD Infinity Context: A Purpose-Built KV Cache Tier for Distributed Inference
AMD · 22 iul. 2026
Understanding Is Done Early: A Depth Division of Labor in Large Language Models and Its Use for Unbounded-Context Memory
arXiv · 31 iul. 2026