{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/cross-self-kv-cache-pruning-for-efficient","title":"Cross-Self KV Cache Pruning for Efficient Vision-Language Inference","arxiv_id":"2412.04652","date":"2024-12-05","proceeding":null,"authors":["Xiaohuan Pei","Tao Huang","Chang Xu"],"abstract":"KV cache pruning has emerged as a promising technique for reducing memory and computation costs in long-context auto-regressive generation. Existing methods for vision-language models (VLMs) typically rely on self-attention scores from large language models (LLMs) to identify and prune irrelevant tokens. However, these approaches overlook the inherent distributional discrepancies between modalities, often leading to inaccurate token importance estimation and the over-pruning of critical visual tokens. To address this, we propose decomposing attention scores into intra-modality attention (within the same modality) and inter-modality attention (across modalities), enabling more precise KV cache pruning by independently managing these distinct attention types. Additionally, we introduce an n-softmax function to counteract distribution shifts caused by pruning, preserving the original smoothness of attention scores and ensuring stable performance. Our final training-free method, \\textbf{C}ross-\\textbf{S}elf \\textbf{P}runing (CSP), achieves competitive performance compared to models with full KV caches while significantly outperforming previous pruning methods. Extensive evaluations on MileBench, a benchmark encompassing 29 multimodal datasets, demonstrate CSP's effectiveness, achieving up to a 41\\% performance improvement on challenging tasks like conversational embodied dialogue while reducing the KV cache budget by 13.6\\%. The code is available at https://github.com/TerryPei/CSP","url_abs":"https://arxiv.org/abs/2412.04652v1","url_pdf":"https://arxiv.org/pdf/2412.04652v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"cross-self-kv-cache-pruning-for-efficient","repo_url":"https://github.com/terrypei/csp","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[],"methods":[{"method_slug":"attention","method_name":"Attention"},{"method_slug":"pruning","method_name":"Pruning"},{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2412.04652","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2412.04652"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/terrypei/csp","reach":null}],"summary":{"ran_fixture":1,"ran_violates":1},"by_repo_kind":{"official":{"samples":1,"ran":1,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":1,"samples":[{"code_sha256_prefix":"30d7eec482ebf6b1","entry":"repeat_kv","repo":null,"repo_kind":null,"path":null,"file_url":null,"link_basis":"identical_code_first_harvested_elsewhere","language":"python","status":"ran_fixture","verification_level":2,"contract_check":"RAISES","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":null,"inline_ok":false,"mcp_get_code":{"code_sha256":"30d7eec482ebf6b1"}},{"code_sha256_prefix":"506a6701a4ac56f7","entry":"kl_balance","repo":"terrypei/csp","repo_kind":"official","path":"LLaVA-mix_merge_v1/llava/model/kv_token_merge/modify_llama.py","file_url":"https://github.com/terrypei/csp/blob/HEAD/LLaVA-mix_merge_v1/llava/model/kv_token_merge/modify_llama.py","link_basis":"first_harvest_node","language":"python","status":"ran_violates","verification_level":1,"contract_check":"VIOLATES","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"506a6701a4ac56f7"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}