Dialogue Without Limits: Constant-Sized KV Caches for Extended Responses in LLMs
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866909642166108160 |
|---|---|
| author | Ghadia, Ravi Kumar, Avinash Jain, Gaurav Nair, Prashant Das, Poulami |
| author_facet | Ghadia, Ravi Kumar, Avinash Jain, Gaurav Nair, Prashant Das, Poulami |
| contents | Autoregressive Transformers rely on Key-Value (KV) caching to accelerate inference. However, the linear growth of the KV cache with context length leads to excessive memory consumption and bandwidth constraints. This bottleneck is particularly problematic in real-time applications -- such as chatbots and interactive assistants -- where low latency and high memory efficiency are critical. Existing methods drop distant tokens or compress states in a lossy manner, sacrificing accuracy by discarding vital context or introducing bias.
We propose MorphKV, an inference-time technique that maintains a constant-sized KV cache while preserving accuracy. MorphKV balances long-range dependencies and local coherence during text generation. It eliminates early-token bias while retaining high-fidelity context by adaptively ranking tokens through correlation-aware selection. Unlike heuristic retention or lossy compression, MorphKV iteratively refines the KV cache via lightweight updates guided by attention patterns of recent tokens. This approach captures inter-token correlation with greater accuracy, crucial for tasks like content creation and code generation. Our studies on long-response tasks show 52.9$\%$ memory savings and 18.2$\%$ higher accuracy on average compared to state-of-the-art prior works, enabling efficient real-world deployment. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_00979 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Dialogue Without Limits: Constant-Sized KV Caches for Extended Responses in LLMs Ghadia, Ravi Kumar, Avinash Jain, Gaurav Nair, Prashant Das, Poulami Computation and Language Artificial Intelligence Machine Learning Autoregressive Transformers rely on Key-Value (KV) caching to accelerate inference. However, the linear growth of the KV cache with context length leads to excessive memory consumption and bandwidth constraints. This bottleneck is particularly problematic in real-time applications -- such as chatbots and interactive assistants -- where low latency and high memory efficiency are critical. Existing methods drop distant tokens or compress states in a lossy manner, sacrificing accuracy by discarding vital context or introducing bias. We propose MorphKV, an inference-time technique that maintains a constant-sized KV cache while preserving accuracy. MorphKV balances long-range dependencies and local coherence during text generation. It eliminates early-token bias while retaining high-fidelity context by adaptively ranking tokens through correlation-aware selection. Unlike heuristic retention or lossy compression, MorphKV iteratively refines the KV cache via lightweight updates guided by attention patterns of recent tokens. This approach captures inter-token correlation with greater accuracy, crucial for tasks like content creation and code generation. Our studies on long-response tasks show 52.9$\%$ memory savings and 18.2$\%$ higher accuracy on average compared to state-of-the-art prior works, enabling efficient real-world deployment. |
| title | Dialogue Without Limits: Constant-Sized KV Caches for Extended Responses in LLMs |
| topic | Computation and Language Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2503.00979 |