Pith. sign in

REVIEW 8 cited by

Dynamic Memory Compression: Retrofitting LLMs for Accelerated Inference

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2403.09636 v2 pith:UPKZGQLX submitted 2024-03-14 cs.CL

Dynamic Memory Compression: Retrofitting LLMs for Accelerated Inference

classification cs.CL
keywords compressionllmsmemorycacheinferencekey-valuedifferentdynamic
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Transformers have emerged as the backbone of large language models (LLMs). However, generation remains inefficient due to the need to store in memory a cache of key-value representations for past tokens, whose size scales linearly with the input sequence length and batch size. As a solution, we propose Dynamic Memory Compression (DMC), a method for online key-value cache compression at inference time. Most importantly, the model learns to apply different compression ratios in different heads and layers. We retrofit pre-trained LLMs such as Llama 2 (7B, 13B and 70B) into DMC Transformers, achieving up to 7x throughput increase during auto-regressive inference on an NVIDIA H100 GPU. DMC is applied via continued pre-training on a negligible percentage of the original data without adding any extra parameters. DMC preserves the original downstream performance with up to 4x cache compression, outperforming up-trained grouped-query attention (GQA) and key-value eviction policies (H$_2$O, TOVA). GQA and DMC can be even combined to obtain compounded gains. Hence, DMC can serve as a drop-in replacement for KV caching in existing LLMs to fit longer contexts and larger batches within any given memory budget.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 8 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Sparse Prefix Caching for Hybrid and Recurrent LLM Serving

    cs.LG 2026-04 unverdicted novelty 7.0

    Sparse prefix caching via dynamic programming for optimal checkpoint placement under overlap distributions improves the Pareto frontier for recurrent and hybrid LLM serving on shared-prefix data.

  2. What to Keep, What to Forget: A Rate--Distortion View of Memory Compaction in LLMs and Agents

    cs.LG 2026-07 conditional novelty 6.0

    KV-cache eviction, prompt compression, recurrent state bounding, and agent memory consolidation are unified as one rate-distortion problem with a shared lower bound, shared failure mode, and transferable mechanisms.

  3. MomentKV: Closing the Directional Gap in KV Cache Eviction for Long-Context Inference

    cs.LG 2026-06 unverdicted novelty 6.0

    MomentKV maintains count, key mean, value mean, and value-key covariance over evicted tokens to guide selective eviction and provide a first-order approximation of their attention contribution, outperforming baselines...

  4. Make Each Token Count: Towards Improving Long-Context Performance with KV Cache Eviction

    cs.LG 2026-05 unverdicted novelty 6.0

    A unified learnable KV eviction policy with cross-layer calibration reduces memory and matches or exceeds full-cache performance on long-context tasks by retaining useful tokens and limiting attention dilution.

  5. eOptShrinkQ: Near-Lossless KV Cache Compression Through Optimal Spectral Denoising and Quantization

    cs.LG 2026-04 unverdicted novelty 6.0

    eOptShrinkQ compresses KV caches to ~2.2 bits per entry via optimal spectral shrinkage and quantization, outperforming prior methods on LongBench while matching FP16 on multi-needle retrieval.

  6. LightTransfer: Your Long-Context LLM is Secretly a Hybrid Model with Effortless Adaptation

    cs.CL 2024-10 unverdicted novelty 6.0

    LightTransfer identifies lazy layers in LLMs like LLaMA and replaces their attention with streaming attention to form hybrid models, delivering up to 2.17x throughput with under 1.5% drop on LongBench and strong resul...

  7. PaceLLM: Brain-Inspired Large Language Models for Long-Context Understanding

    q-bio.NC 2025-06 unverdicted novelty 5.0

    PaceLLM introduces brain-inspired persistent activity and cortical expert clustering to mitigate contextual decay and semantic fragmentation in LLMs, reporting 6% gains on LongBench multi-document QA and 12.5-17.5% on...

  8. A Survey of Mamba

    cs.LG 2024-08 unverdicted novelty 2.0

    The paper consolidates existing research on Mamba models, their architecture variants, adaptations to different data modalities, and applications across domains.