{"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/arxiv-2606-06302","title":"Tangram: Unlocking Non-Uniform KV Cache Compression for Efficient Multi-turn LLM Serving","arxiv_id":"2606.06302","date":"2026-06-04","proceeding":null,"authors":["Hyungmin Kim","Minsoo Kim","Hongseok Kim","Jungwook Choi"],"abstract":"Multi-turn LLM serving accumulates dialogue history whose Key-Value (KV) cache grows with every turn and every user, quickly exceeding the model weights themselves and making memory -- not compute -- the binding constraint on throughput. Non-uniform KV compression, which allocates heterogeneous budgets across attention heads, preserves accuracy far better than uniform schemes, yet remains impractical: modern serving stacks assume identical KV lengths across heads, so heterogeneity traps freed memory as page fragmentation, spends up to 25% of prefill time reclaiming scattered pages, and skews GPU workloads that inflate decode latency by up to $1.7\\times$ or burn 15--20% of each decode step on re-planning. We observe that this heterogeneity need not be discovered at runtime: head-wise retention follows a two-level structural regularity -- an input-invariant head ranking with narrowly bounded per-head ratios -- that can be calibrated offline from as few as 50 samples. Building on this insight, we present Tangram, a serving framework that statically resolves what prior systems handle dynamically: Budget Reservation fixes each head's post-compression footprint at scheduling time, eliminating page reclamation; Ragged Paging clusters similar-budget heads into independent page tables, turning fragmentation into reclaimable memory; and Ahead-of-Time Load Balancing precomputes balanced GPU partitions with zero runtime planning. Implemented on vLLM, Tangram serves as a drop-in substrate for existing non-uniform compression methods, matching their accuracy while improving end-to-end throughput by up to $2.6\\times$ over the full-KV baseline. Our implementation is publicly available at https://github.com/aiha-lab/TANGRAM.","url_abs":"https://arxiv.org/abs/2606.06302","url_pdf":"https://arxiv.org/pdf/2606.06302","source":{"archive":null,"snapshot":"2025-07-28","note":"not in the Papers with Code archive (frozen at the snapshot)","row_kind":"graph","title_abstract_authors_date":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)"},"code_links":[],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2606.06302","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2606.06302"}},"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":[{"mentioned_in_github":null,"is_official":null,"provenance":"deterministic:regex_extraction","mentioned_in_paper":null,"url":"https://github.com/aiha-lab/TANGRAM","reach":null}],"summary":{"ran":12,"unverified":5},"by_repo_kind":{"found_in_text":{"samples":17,"ran":12,"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":0,"samples":[{"code_sha256_prefix":"552e4370076a1a01","entry":"async_request_deepspeed_mii","repo":"aiha-lab/TANGRAM","repo_kind":"found_in_text","path":"benchmarks/backend_request_func.py","file_url":"https://github.com/aiha-lab/TANGRAM/blob/HEAD/benchmarks/backend_request_func.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"552e4370076a1a01"}},{"code_sha256_prefix":"7ee13b8d142453c0","entry":"async_request_tgi","repo":"aiha-lab/TANGRAM","repo_kind":"found_in_text","path":"benchmarks/backend_request_func.py","file_url":"https://github.com/aiha-lab/TANGRAM/blob/HEAD/benchmarks/backend_request_func.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"7ee13b8d142453c0"}},{"code_sha256_prefix":"f0b0f51ad4ed00c9","entry":"async_request_trt_llm","repo":"aiha-lab/TANGRAM","repo_kind":"found_in_text","path":"benchmarks/backend_request_func.py","file_url":"https://github.com/aiha-lab/TANGRAM/blob/HEAD/benchmarks/backend_request_func.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"f0b0f51ad4ed00c9"}},{"code_sha256_prefix":"bcd9d69939662364","entry":"build_prompt","repo":"aiha-lab/TANGRAM","repo_kind":"found_in_text","path":"benchmarks/tangram/ruler_local.py","file_url":"https://github.com/aiha-lab/TANGRAM/blob/HEAD/benchmarks/tangram/ruler_local.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"bcd9d69939662364"}},{"code_sha256_prefix":"4d80b1ddfa08469e","entry":"effective_ratio","repo":"aiha-lab/TANGRAM","repo_kind":"found_in_text","path":"benchmarks/tangram/bench_common.py","file_url":"https://github.com/aiha-lab/TANGRAM/blob/HEAD/benchmarks/tangram/bench_common.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"4d80b1ddfa08469e"}},{"code_sha256_prefix":"418bfc593ad8d8bf","entry":"extract_request_timing","repo":"aiha-lab/TANGRAM","repo_kind":"found_in_text","path":"benchmarks/tangram/bench_common.py","file_url":"https://github.com/aiha-lab/TANGRAM/blob/HEAD/benchmarks/tangram/bench_common.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"418bfc593ad8d8bf"}},{"code_sha256_prefix":"550815cfa153d37c","entry":"get_data_list","repo":"aiha-lab/TANGRAM","repo_kind":"found_in_text","path":"benchmarks/tangram/scbench_local.py","file_url":"https://github.com/aiha-lab/TANGRAM/blob/HEAD/benchmarks/tangram/scbench_local.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"550815cfa153d37c"}},{"code_sha256_prefix":"6439d722e41feedf","entry":"get_eval_task","repo":"aiha-lab/TANGRAM","repo_kind":"found_in_text","path":"benchmarks/tangram/benchmark_scbench.py","file_url":"https://github.com/aiha-lab/TANGRAM/blob/HEAD/benchmarks/tangram/benchmark_scbench.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"6439d722e41feedf"}},{"code_sha256_prefix":"941d33a97c872155","entry":"get_query_text","repo":"aiha-lab/TANGRAM","repo_kind":"found_in_text","path":"benchmarks/tangram/benchmark_scbench.py","file_url":"https://github.com/aiha-lab/TANGRAM/blob/HEAD/benchmarks/tangram/benchmark_scbench.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"941d33a97c872155"}},{"code_sha256_prefix":"d59e2d450cbbf7a4","entry":"group_by_task","repo":"aiha-lab/TANGRAM","repo_kind":"found_in_text","path":"benchmarks/tangram/ruler_local.py","file_url":"https://github.com/aiha-lab/TANGRAM/blob/HEAD/benchmarks/tangram/ruler_local.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"d59e2d450cbbf7a4"}},{"code_sha256_prefix":"ac80359234193760","entry":"set_gen_length","repo":"aiha-lab/TANGRAM","repo_kind":"found_in_text","path":"benchmarks/tangram/scbench_local.py","file_url":"https://github.com/aiha-lab/TANGRAM/blob/HEAD/benchmarks/tangram/scbench_local.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"ac80359234193760"}},{"code_sha256_prefix":"33321941cb7c5a2a","entry":"template","repo":"aiha-lab/TANGRAM","repo_kind":"found_in_text","path":"benchmarks/tangram/scbench_local.py","file_url":"https://github.com/aiha-lab/TANGRAM/blob/HEAD/benchmarks/tangram/scbench_local.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"33321941cb7c5a2a"}},{"code_sha256_prefix":"42a1756632040233","entry":"load_ruler","repo":"aiha-lab/TANGRAM","repo_kind":"found_in_text","path":"benchmarks/tangram/ruler_local.py","file_url":"https://github.com/aiha-lab/TANGRAM/blob/HEAD/benchmarks/tangram/ruler_local.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"42a1756632040233"}},{"code_sha256_prefix":"f942346fd9105175","entry":"resolve_cluster_map","repo":"aiha-lab/TANGRAM","repo_kind":"found_in_text","path":"benchmarks/tangram/speedup/speedup_benchmark.py","file_url":"https://github.com/aiha-lab/TANGRAM/blob/HEAD/benchmarks/tangram/speedup/speedup_benchmark.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"f942346fd9105175"}},{"code_sha256_prefix":"db377885d2303bb9","entry":"run_config_in_subprocess","repo":"aiha-lab/TANGRAM","repo_kind":"found_in_text","path":"benchmarks/tangram/preemption_overhead/benchmark.py","file_url":"https://github.com/aiha-lab/TANGRAM/blob/HEAD/benchmarks/tangram/preemption_overhead/benchmark.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"db377885d2303bb9"}},{"code_sha256_prefix":"2298f1ff53dfe660","entry":"run_one","repo":"aiha-lab/TANGRAM","repo_kind":"found_in_text","path":"benchmarks/tangram/preemption_overhead/benchmark.py","file_url":"https://github.com/aiha-lab/TANGRAM/blob/HEAD/benchmarks/tangram/preemption_overhead/benchmark.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"2298f1ff53dfe660"}},{"code_sha256_prefix":"8fb5af6a89b21e9b","entry":"run_one","repo":"aiha-lab/TANGRAM","repo_kind":"found_in_text","path":"benchmarks/tangram/speedup/speedup_benchmark.py","file_url":"https://github.com/aiha-lab/TANGRAM/blob/HEAD/benchmarks/tangram/speedup/speedup_benchmark.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"8fb5af6a89b21e9b"}}]},"arxiv_metadata":{"licence":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)","fields":["title","abstract","authors","date"],"primary_category":"cs.LG","source":"arxiv_2026.jsonl"},"syntology_extracted_results":{"kind":"leaderboard_placements","source":"Syntology's leaderboard-shaped extractor over the paper's own arXiv-HTML tables: a model pointed at a cell, the number was read from that cell and checked against the board's metric, dataset, split and scale, and an independent check accepted the entry; not reviewed by the paper's authors or the archive's editors","extractor_model":"global.anthropic.claude-sonnet-4-5-20250929-v1:0","verifier_model":null,"prompt_sha":"fa63d4bb9d755694","coverage":{"sentence":"Syntology has checked 6,264 of the 9,581 papers on this site that are newer than the archive; results from the others appear after they are checked.","papers_newer_than_archive":9581,"papers_checked":6264},"entries":[],"not_placed":{"boards":0,"rejected_by_independent_check":0,"refused_by_a_rule":0,"check_did_not_answer":0,"proposed_without_a_cell":0,"declined_by_site":0}}}