Organizations: Georgia Institute of Technology, USA · Microsoft; work done while at Georgia Institute of Technology. · Microsoft; work done while he was employed by Cisco Systems.
LLM inference has two phases with different bottlenecks: LLM prefill runs once per request and is compute-intensive. Decoding repeats for every output token and is limited by memory bandwidth. To address the challenges associated with this issue, different solutions have been proposed. All of these solutions use the same model for both prefill and decode phases. We propose BridgeKV, which uses a larger model for prefill and a smaller model of the same family for decoding. A lightweight KV-stitching module translates the larger model's prompt cache into the smaller model's cache layout. It handles differences in cache shape and representation through learned local layer mixing and separate nonlinear maps for keys and values. With both models frozen, we train the module to align their caches and make the smaller model follow the larger model's next-token predictions. A second training stage adds low-rank adapters when further task adaptation is needed. Across six model pairs and five datasets, BridgeKV outperforms fine-tuned small-model baselines with matched trainable parameter budgets in all 30 settings. It closes up to 80% of the quality gap to using the larger model for both phases. On one A40 GPU, Llama-3.1-Instruct-8B to Llama-3.2-Instruct-1B achieves 3.0X faster requests than the larger model alone, with latency close to the smaller model alone. BridgeKV can improve the hardware and system solutions that separate prefill and decode phases across GPUs. A separate serving simulation models request arrivals, queues, and cache transfer. For Llama-3.1-Instruct-70B to Llama-3.2-Instruct-1B, it predicts up to 33.86X faster requests and 27.5X lower energy use than using the larger model for both prefill and decode phases.
Multi-model LLM serving uses routers, cascades and agent workflows to move requests across models with different costs and capabilities. Existing serving systems reuse KV caches only within one model, so every switch forces the target to re-prefill history the source already processed, adding time to first token (TTFT) that grows with history length and target size. Cross-model KV transfer can replace this prefill, yet prior methods are slower than re-prefill end to end. We present CacheBridge, the first serving system to accelerate heterogeneous multi-model workloads through cross-model KV transfer. It co-designs a compact, attention-calibrated KV mapper with systematic optimization for online multi-model serving, which makes model switch efficient on latency and memory cost. In Qwen3 14B→32B agent serving on τ-bench, it reduces mean TTFT by 30 to 98% versus re-prefill at 1 to 24 QPS with comparable task success. Across five model pairs from three families, it uses 8 to 12× less mapper memory than Closed-Form Mapper and retains on average 97.94% and 102.94% of re-prefill task quality on chat/reasoning and RAG workloads, respectively. The promising results have opened new opportunities for efficient multi-model LLM serving through cross-model KV reuse.
The large size of the KV cache has become a major bottleneck for serving LLMs with increasing context lengths. In response, many KV cache compression methods, such as token dropping and quantization, have been proposed. However, almost all of these methods are inherently lossy-despite minimal accuracy degradation for short outputs, their outputs increasingly diverge from full-KV-cache outputs as more tokens are decoded, which leads to catastrophic failures in code generation and tool calling. We present VeriCache, the first inference framework that ensures the same output as full-KV-cache decoding but largely preserves the high decoding throughput of a range of KV cache compression algorithms. VeriCache uses the compressed KV cache to draft tokens, then verifies them against the full KV cache. While it may seem like just speculative decoding, VeriCache requires addressing a key system challenge to work-keeping the full KV cache out of GPU memory and minimizing the overhead of swapping it in for verification. The insight is two-fold: (1) compressed-KV decoding can be parallelized with full-KV swap, because one is HBM-bandwidth-bound and the other is PCIe/network-bound, and (2) the compressed KV cache often produces output similar to the full KV cache, allowing a long drafting horizon to amortize each full-KV swap. VeriCache applies to both long-context decoding and remote prefix caching, supports a broad family of token-dropping and quantization methods through a uniform compressor interface, and composes with traditional speculative decoding. Experimental results show that VeriCache achieves up to 4X higher throughput than full-KV inference while producing identical outputs.
Jiayi Yao, Samuel Shen, Kuntai Du +7
University of Chicago · Tensormesh Inc. · Samsung Semiconductor +1
Cross-model KV-cache reuse remains a key challenge in modern LLM serving. Coding agents and multi-model systems increasingly route a shared context across models: a user may switch models mid-session, or a cascade may escalate a difficult query. Because KV caches contain model-specific representations, each switch typically forces the receiving model to prefill the entire context from scratch. Recent work shows that closed-form linear maps can translate KV caches between models in the same family, but transfer accuracy degrades as the model-size gap widens. In this paper, we establish that these transfer failures are concentrated in a small subset of information-dense tokens. To bridge this gap, we introduce RaReCache, a framework that enables a large target model to decode accurately from a cache prefilled by a much smaller source via selective recomputation. RaReCache identifies these critical positions using a novel rank disagreement metric, scoring each token by the energy of its mapped KV in output directions weakly supported by the calibration data. Across two model families and five benchmarks, on a 23x parameter gap (Qwen3-0.6B to 14B) recomputing just 30% of positions retains 95-99% of the target accuracy, whereas on a 8.8x gap (Llama3-8B to 70B), recomputing 40% retains 96.5% of the target accuracy. RaReCache largely removes sensitivity to source-model size, and achieves up to a 3.04x prefill speedup. For online serving, it handles 1.8x the request throughput of target prefill on a single GPU, and at the target's saturation load, reduces median and 99th-percentile time-to-first-token (TTFT) by 5.0x and 6.4x respectively, with a 30% recompute budget. RaReCache establishes an efficient serving paradigm where small models prefill on behalf of massive targets, enabling large models to recompute only critical tokens, drastically reducing prefill latency.
Sreetama Sarkar, Saptarshi Mitra, Sitao Huang +2
University of Southern California · University of California, Irvine · Intel Labs