{"about":{"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.","site":"https://codewithpapers.app","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","syntology":{"site":"https://syntology.ai","developers":"https://syntology.ai/developers","mcp":{"server":"https://syntology.ai/mcp","transport":"streamable-http","server_card":"https://syntology.ai/.well-known/mcp/server-card.json","auth":{"type":"trial token, no account","trial_token":"https://syntology.ai/api/oauth/trial/token","method":"POST","docs":"https://syntology.ai/developers"}},"have":"https://syntology.ai/api/graph/have?x=<method, arXiv id or title> (free, answers coverage only)","paper_base":"https://syntology.ai/paper/","atlas_base":"https://app.syntology.ai/?focus="},"machine_readable":[{"url":"https://codewithpapers.app/llms.txt","what":"the machine catalog: every machine-readable file, counted"},{"url":"https://codewithpapers.app/index/manifest.json","what":"paper-to-code index by arXiv id, with Syntology's counts"},{"url":"https://codewithpapers.app/search/manifest.json","what":"site search index (titles, authors) and its files"},{"url":"https://codewithpapers.app/download","what":"bulk files: Syntology's layer, described there"},{"url":"https://codewithpapers.app/build_manifest.json","what":"the build record: inputs, counts, exclusions, probes"}]},"url":"/method/attention-sinks","slug":"attention-sinks","name":"Attention Sinks","full_name":"Attention Sinks","full_name_withheld":false,"description_markdown":null,"description_state":"placeholder","introduced_year":null,"introduced_by":{"title":"Efficient Streaming Language Models with Attention Sinks","paper":"/paper/efficient-streaming-language-models-with","first_author":"Guangxuan Xiao","n_authors":5,"url_abs":null,"archive_paper_url":"https://paperswithcode.com/paper/efficient-streaming-language-models-with"},"source":{"url":"https://arxiv.org/abs/2309.17453v4","title":"Efficient Streaming Language Models with Attention Sinks","url_on_a_paper_host":true},"code_snippet_url":null,"code_snippet_url_on_a_code_host":false,"categories":[{"area":"General","area_id":"general","collection":"Attention","url":"/methods/category/attention","pwc_aliases":[]}],"n_papers_tagged":13,"archive_num_papers":13,"papers_newest_first":[{"paper":null,"slug":"mars-bench-a-multi-turn-athletic-real-world","title":"MARS-Bench: A Multi-turn Athletic Real-world Scenario Benchmark for Dialogue Evaluation","date":"2025-05-27","arxiv_id":"2505.23810","n_code_links":0,"syntology":null},{"paper":null,"slug":"analysis-of-attention-in-video-diffusion","title":"Analysis of Attention in Video Diffusion Transformers","date":"2025-04-14","arxiv_id":"2504.10317","n_code_links":0,"syntology":null},{"paper":"/paper/why-do-llms-attend-to-the-first-token","slug":"why-do-llms-attend-to-the-first-token","title":"Why do LLMs attend to the first token?","date":"2025-04-03","arxiv_id":"2504.02732","n_code_links":1,"syntology":{"ran":5,"of":7,"n_ran_checked":5,"n_instrument":0,"unverified":2,"pointer_only":0,"phrase":"5 ran (of which 0 constructed an object rather than computing a result; 5 with no instrument failure: 0 honoured, 0 violated, 5 with no contract checked; 0 where Syntology's instrument failed) · 2 unverified","official":null}},{"paper":"/paper/interpreting-the-repeated-token-phenomenon-in","slug":"interpreting-the-repeated-token-phenomenon-in","title":"Interpreting the Repeated Token Phenomenon in Large Language Models","date":"2025-03-11","arxiv_id":"2503.08908","n_code_links":1,"syntology":null},{"paper":null,"slug":"attention-sinks-and-outlier-features-a-catch","title":"Attention Sinks and Outlier Features: A 'Catch, Tag, and Release' Mechanism for Embeddings","date":"2025-02-02","arxiv_id":"2502.00919","n_code_links":0,"syntology":null},{"paper":null,"slug":"task-kv-task-aware-kv-cache-optimization-via","title":"Task-KV: Task-aware KV Cache Optimization via Semantic Differentiation of Attention Heads","date":"2025-01-25","arxiv_id":"2501.15113","n_code_links":0,"syntology":null},{"paper":null,"slug":"attention-entropy-is-a-key-factor-an-analysis","title":"Attention Entropy is a Key Factor: An Analysis of Parallel Context Encoding with Full-attention-based Pre-trained Language Models","date":"2024-12-21","arxiv_id":"2412.16545","n_code_links":0,"syntology":null},{"paper":null,"slug":"seeing-clearly-by-layer-two-enhancing","title":"Seeing Clearly by Layer Two: Enhancing Attention Heads to Alleviate Hallucination in LVLMs","date":"2024-11-15","arxiv_id":"2411.09968","n_code_links":0,"syntology":null},{"paper":"/paper/value-residual-learning-for-alleviating","slug":"value-residual-learning-for-alleviating","title":"Value Residual Learning For Alleviating Attention Concentration In Transformers","date":"2024-10-23","arxiv_id":"2410.17897","n_code_links":1,"syntology":{"ran":9,"of":12,"n_ran_checked":4,"n_instrument":5,"unverified":3,"pointer_only":1,"phrase":"9 ran (of which 0 constructed an object rather than computing a result; 4 with no instrument failure: 0 honoured, 0 violated, 4 with no contract checked; 5 where Syntology's instrument failed) · 3 unverified","official":{"repos":["Zcchill/Value-Residual-Learning"],"state":"official (archive's flag): 9 ran","n_ran":9,"n_constructed":0,"n_ran_no_instrument_failure":4,"n_unverified":3,"ran_from_kinds":["official"]}}},{"paper":"/paper/when-attention-sink-emerges-in-language","slug":"when-attention-sink-emerges-in-language","title":"When Attention Sink Emerges in Language Models: An Empirical View","date":"2024-10-14","arxiv_id":"2410.10781","n_code_links":1,"syntology":{"ran":3,"of":3,"n_ran_checked":2,"n_instrument":1,"unverified":0,"pointer_only":0,"phrase":"3 ran (of which 0 constructed an object rather than computing a result; 2 with no instrument failure: 2 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified","official":{"repos":["sail-sg/attention-sink"],"state":"official (archive's flag): 3 ran","n_ran":3,"n_constructed":0,"n_ran_no_instrument_failure":2,"n_unverified":0,"ran_from_kinds":["official"]}}},{"paper":"/paper/does-roberta-perform-better-than-bert-in","slug":"does-roberta-perform-better-than-bert-in","title":"Does RoBERTa Perform Better than BERT in Continual Learning: An Attention Sink Perspective","date":"2024-10-08","arxiv_id":"2410.05648","n_code_links":1,"syntology":null},{"paper":"/paper/unveiling-and-harnessing-hidden-attention","slug":"unveiling-and-harnessing-hidden-attention","title":"Unveiling and Harnessing Hidden Attention Sinks: Enhancing Large Language Models without Training through Attention Calibration","date":"2024-06-22","arxiv_id":"2406.15765","n_code_links":1,"syntology":{"ran":5,"of":13,"n_ran_checked":1,"n_instrument":4,"unverified":8,"pointer_only":13,"phrase":"5 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 4 where Syntology's instrument failed) · 8 unverified","official":{"repos":["gatech-eic/act"],"state":"official (archive's flag): 5 ran","n_ran":5,"n_constructed":0,"n_ran_no_instrument_failure":1,"n_unverified":8,"ran_from_kinds":["official"]}}},{"paper":"/paper/efficient-streaming-language-models-with","slug":"efficient-streaming-language-models-with","title":"Efficient Streaming Language Models with Attention Sinks","date":"2023-09-29","arxiv_id":"2309.17453","n_code_links":6,"syntology":{"ran":8,"of":11,"n_ran_checked":4,"n_instrument":4,"unverified":3,"pointer_only":2,"phrase":"8 ran (of which 4 constructed an object rather than computing a result; 4 with no instrument failure: 0 honoured, 0 violated, 4 with no contract checked; 4 where Syntology's instrument failed) · 3 unverified","official":{"repos":["intel/intel-extension-for-transformers","mit-han-lab/streaming-llm"],"state":"official (archive's flag): 6 ran","n_ran":6,"n_constructed":2,"n_ran_no_instrument_failure":2,"n_unverified":0,"ran_from_kinds":["listed","official"]}}}],"papers_shown":13,"papers_order":"date (newest first), then slug","paper_counts":{"archive_papers_tagged":13,"with_a_code_link":7,"where_syntology_ran_a_sample":5,"not_listed_spam_title":0,"listed":13,"listed_where_code_ran":5,"where_syntology_ran_a_sample_split":{"with_a_run_with_no_instrument_failure":5,"every_run_a_failure_of_syntologys_instrument":0,"listed_with_a_run_with_no_instrument_failure":5,"listed_every_run_a_failure_of_syntologys_instrument":0,"filter":{"states":["a run with no instrument failure","any run, instrument failures included"],"default":"a run with no instrument failure","note":"on the 'only where code ran' pages the default hides, in the browser, the rows where every run was a failure of Syntology's instrument; the second state shows them again. Rows are hidden, never re-ordered; these twins list every row"}},"definition":"distinct papers the archive tags; 'where Syntology ran a sample' counts papers with at least one harvested sample that ran, which is not a correctness claim"},"papers_where_code_ran":5,"papers_where_code_ran_first_page":"/method/attention-sinks/papers/ran/1","papers_where_code_ran_caption":"We ran code from the paper's repository; we did not isolate this method inside it.","pages":[{"order":"ran","page":1,"of":1,"url":"/method/attention-sinks/papers/ran/1","json":"/method/attention-sinks/papers/ran/1.json","rows":5}],"tasks":[{"task":"/task/continual-learning","name":"Continual Learning","papers":1},{"task":"/task/decoder","name":"Decoder","papers":1},{"task":"/task/dialogue-evaluation","name":"Dialogue Evaluation","papers":1},{"task":"/task/diversity","name":"Diversity","papers":1},{"task":"/task/hallucination","name":"Hallucination","papers":1},{"task":"/task/language-modeling","name":"Language Modeling","papers":1},{"task":"/task/language-modelling","name":"Language Modelling","papers":1},{"task":"/task/model-compression","name":"Model Compression","papers":1},{"task":"/task/quantization","name":"Quantization","papers":1},{"task":"/task/tag","name":"TAG","papers":1},{"task":"/task/video-editing","name":"Video Editing","papers":1}],"tasks_shown":11,"n_tasks":11,"usage_by_year":[{"year":"2023","papers":1},{"year":"2024","papers":6},{"year":"2025","papers":6}],"row_source":"methods_table","archive":{"source":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","archive_url":"https://paperswithcode.com/method/attention-sinks"},"syntology_read_at":"2026-10-05T00:36:11+00:00","syntology":{"source":"Syntology's have check for this name (the free coverage lookup), read from Syntology's graph","have_url":"https://syntology.ai/api/graph/have?x=Attention%20Sinks","read_at":"2026-10-05T00:52:47+00:00","query":"Attention Sinks","kind":"paper","have":"unknown_name","tier":null,"verified_implementation":null,"level":null,"level_meaning":"the served verification level, per answer; never summed across methods","resolved_to":null,"resolution":null,"resolved_name_differs":false,"absence_asserted":false,"absence_unchecked":null,"no_code_is_about_this_method":false,"subject":null,"archive_paper_arxiv_id":"2309.17453","repositories":null,"retrieve_with":null,"name_competition":{"verdict":"unresolved_bearers","may_be_wrong_sense":false,"bearers_found":2,"served":null,"leading_other_bearer":{"arxiv_id":"2502.00919","title":"Attention Sinks: A 'Catch, Tag, Release' Mechanism for Embeddings","citations":13,"named_by":["title_prefix"],"bearer_name":"Attention Sinks: A 'Catch, Tag, Release' Mechanism for Embeddings","ask_with":"have('2502.00919')"},"n_other_bearers_listed":2,"note":"Nothing resolved to 'Attention Sinks' by name, but 2 thing(s) here DECLARE that name -- most cited: the paper \"Attention Sinks: A 'Catch, Tag, Release' Mechanism for Embeddings\" (arXiv:2502.00919, 13 citations). None of them is a well-cited bearer of the name, so this is still a coverage gap for any better-known sense -- they are listed so you can judge, not as a substitute."},"name_competition_text":"Nothing resolved to this name, but 2 entries in Syntology's graph carry it; the most cited is the paper “Attention Sinks: A 'Catch, Tag, Release' Mechanism for Embeddings” (arXiv:2502.00919, 13 citations). None of them is well cited, so for any better known sense of the name this is still a gap in coverage; they are listed so you can judge, not as a substitute.","near_matches":["SS-Attention","TF-ATTENTION","VC-Attention","TC-ATTENTION","EL-Attention","Attention-MoA","DFC attention","max-attention"],"code_nearby":[],"code_nearby_caption":null,"origin_paper":null,"line":"name not resolved (unknown_name). Other entries in Syntology's graph carry this name; the lookup's note below says what that means for this name.","level_legend":null},"syntology_have_state":"loaded","record_sha256":"9835683c9550431052394c39d33edc93dc605e21f305f1105304cbb97a0996d4","record_changed_at":"2026-10-05","record_changed_at_basis":"changed"}