{"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-2604-10900","title":"CASK: Core-Aware Selective KV Compression for Reasoning Traces","arxiv_id":"2604.10900","date":"2026-04-13","proceeding":null,"authors":["Buseong Kim","Heejun Gwon"],"abstract":"In large language models performing long-form reasoning, the KV cache grows rapidly with decode length, creating bottlenecks in memory and inference stability. Existing reasoning-oriented KV compression has mostly followed an eviction-centered view: estimate token importance more accurately, then discard lower-ranked entries. Our analysis suggests that scorer refinement alone often fails to substantially reorganize the actual keep-set and may therefore not be the main lever for preserving reasoning behavior. We instead frame reasoning KV compression as a behavior-preserving structured consolidation problem. CASK partitions the decode-time reasoning trace into a protected core that anchors answer formation and intermediate state, and mergeable scratch with high redundancy. The core is preserved, while selective consolidation is applied only to the scratch. To address prompt-heavy regimes where the prefix can exhaust the budget before decode-stage compression becomes active, CASK further uses a two-stage design: prefix eviction followed by decode-stage consolidation. On the H100 reasoning gate, CASK shows higher full-KV continuation fidelity than TriAttention at matched budgets on both AIME24 and AIME25, with recurring cask@384 > triattention@512 crossings. In prompt-heavy replay, multi_news and vcsum act as decode-active witnesses, while qmsum and gov_report expose the prefix_budget_exhausted boundary. The overall evidence supports a simple conclusion: effective reasoning KV compression depends less on more elaborate scorer engineering than on combining core preservation with selective scratch consolidation to lower the usable budget frontier.","url_abs":"https://arxiv.org/abs/2604.10900","url_pdf":"https://arxiv.org/pdf/2604.10900","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=2604.10900","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2604.10900"}},"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/THUDM/LongBench","reach":null}],"summary":{"ran_fixture":1,"ran_honours":1,"ran_draft_wrong":2,"ran_violates":1,"unverified":9},"by_repo_kind":{"found_in_text":{"samples":14,"ran":5,"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":"bc7a7a194c238080","entry":"add_embeddings","repo":"THUDM/LongBench","repo_kind":"found_in_text","path":"LongBench/retrieval/contriever/passage_retrieval.py","file_url":"https://github.com/THUDM/LongBench/blob/HEAD/LongBench/retrieval/contriever/passage_retrieval.py","link_basis":"harvester_set","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"bc7a7a194c238080"}},{"code_sha256_prefix":"b349b79d9cc2934b","entry":"count_score","repo":"THUDM/LongBench","repo_kind":"found_in_text","path":"LongBench/metrics.py","file_url":"https://github.com/THUDM/LongBench/blob/HEAD/LongBench/metrics.py","link_basis":"plan_row","language":"python","status":"ran_honours","verification_level":1,"contract_check":"HONOURS","metamorphic_tier":"well_formed","behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"b349b79d9cc2934b"}},{"code_sha256_prefix":"e7e75981cb464788","entry":"normalize_answer","repo":"THUDM/LongBench","repo_kind":"found_in_text","path":"LongBench/metrics.py","file_url":"https://github.com/THUDM/LongBench/blob/HEAD/LongBench/metrics.py","link_basis":"harvester_set","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"e7e75981cb464788"}},{"code_sha256_prefix":"8c5c581f9264c810","entry":"normalize_zh_answer","repo":"THUDM/LongBench","repo_kind":"found_in_text","path":"LongBench/metrics.py","file_url":"https://github.com/THUDM/LongBench/blob/HEAD/LongBench/metrics.py","link_basis":"plan_row","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"8c5c581f9264c810"}},{"code_sha256_prefix":"4489113b536ca6eb","entry":"post_process","repo":"THUDM/LongBench","repo_kind":"found_in_text","path":"LongBench/pred.py","file_url":"https://github.com/THUDM/LongBench/blob/HEAD/LongBench/pred.py","link_basis":"plan_row","language":"python","status":"ran_violates","verification_level":1,"contract_check":"VIOLATES","metamorphic_tier":"deterministic","behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"4489113b536ca6eb"}},{"code_sha256_prefix":"324506640e971548","entry":"build_chat","repo":"THUDM/LongBench","repo_kind":"found_in_text","path":"LongBench/pred.py","file_url":"https://github.com/THUDM/LongBench/blob/HEAD/LongBench/pred.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"324506640e971548"}},{"code_sha256_prefix":"2e148fc130561394","entry":"build_chat","repo":"THUDM/LongBench","repo_kind":"found_in_text","path":"LongBench/retrieval/pred.py","file_url":"https://github.com/THUDM/LongBench/blob/HEAD/LongBench/retrieval/pred.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"2e148fc130561394"}},{"code_sha256_prefix":"04b0e165da9a0405","entry":"build_chat","repo":"THUDM/LongBench","repo_kind":"found_in_text","path":"LongBench/summ/compress.py","file_url":"https://github.com/THUDM/LongBench/blob/HEAD/LongBench/summ/compress.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"04b0e165da9a0405"}},{"code_sha256_prefix":"7de062709da87b4f","entry":"get_pred","repo":"THUDM/LongBench","repo_kind":"found_in_text","path":"LongBench/retrieval/pred.py","file_url":"https://github.com/THUDM/LongBench/blob/HEAD/LongBench/retrieval/pred.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"7de062709da87b4f"}},{"code_sha256_prefix":"03bffbf21c24cb7b","entry":"get_word_len","repo":"THUDM/LongBench","repo_kind":"found_in_text","path":"LongBench/retrieval/splitter.py","file_url":"https://github.com/THUDM/LongBench/blob/HEAD/LongBench/retrieval/splitter.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"03bffbf21c24cb7b"}},{"code_sha256_prefix":"b80b1e9bf939544f","entry":"get_word_list","repo":"THUDM/LongBench","repo_kind":"found_in_text","path":"LongBench/retrieval/splitter.py","file_url":"https://github.com/THUDM/LongBench/blob/HEAD/LongBench/retrieval/splitter.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"b80b1e9bf939544f"}},{"code_sha256_prefix":"d1a22a0b120644af","entry":"parse_args","repo":"THUDM/LongBench","repo_kind":"found_in_text","path":"LongBench/pred.py","file_url":"https://github.com/THUDM/LongBench/blob/HEAD/LongBench/pred.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"d1a22a0b120644af"}},{"code_sha256_prefix":"58fbc2d5556574d2","entry":"parse_args","repo":"THUDM/LongBench","repo_kind":"found_in_text","path":"LongBench/retrieval/pred.py","file_url":"https://github.com/THUDM/LongBench/blob/HEAD/LongBench/retrieval/pred.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"58fbc2d5556574d2"}},{"code_sha256_prefix":"a4493a19d135ca1d","entry":"split_long_sentence","repo":"THUDM/LongBench","repo_kind":"found_in_text","path":"LongBench/retrieval/splitter.py","file_url":"https://github.com/THUDM/LongBench/blob/HEAD/LongBench/retrieval/splitter.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"a4493a19d135ca1d"}}]},"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}}}