@inproceedings{ramachandran-etal-2026-equip,
    title = "{EQUIP}: {EQU}ivariant preserving In-Place updates for Efficient Token Pruning",
    author = "Ramachandran, Arun  and
      Govindarajan, R  and
      Annavaram, Murali  and
      Raghavendra, Prakash",
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.acl-long.1210/",
    pages = "26303--26323",
    ISBN = "979-8-89176-390-6",
    abstract = "Token-pruning has emerged as a primary focus in large language models (LLMs) to enhance model efficiency while preserving accuracy, especially for large sequence lengths. However, the eviction operation of token-pruning methods causes ``holes'' in KV tensors, posing two major challenges: (1) The shift operation, required to make the KV tensor contiguous, results in significant copy overheads; (2) The changes in position indices due to token eviction lead to increased computational requirements for Rotary Positional Encoding (RoPE). To address these issues, we introduce EQUIP, an EQUivariant preserving in-place token update mechanism that ensures the equivariance property of the operations performed in the attention computation. EQUIP offers two fundamental advantages: First, it combines eviction and a subsequent token insertion into an in-place replacement operation, which reduces the KV cache copy overheads significantly. Second, EQUIP reduces recomputation of rotation operations through a combination of in-place update, caching and a re-indexing strategy. Together, these optimizations enable EQUIP to achieve geomean speedups of 1.62{\texttimes} (or 1.47{\texttimes}) on CPU (GPU) over StreamingLLM, and 3.45{\texttimes} (or 1.86{\texttimes}) on CPU (GPU) over Heavy Hitters (H2O). EQUIP with Paged Attention achieves speedups of 4.18{\texttimes}(2.61{\texttimes}) on CPU (GPU) over auto-regressive baselines. EQUIP matches the model accuracy of baseline pruning methods while delivering superior performance."
}