{"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":"/code/initialize-past-key-values","entry":"initialize_past_key_values","source":"Syntology graph, per-sample; not an archive number","read_at":"2026-10-05T00:36:11+00:00","claim":"Names are grouped by exact entry-name string. Same-named routines are NOT asserted to be equivalent; 'ran' means executed on a synthesized fixture, not correctness. n_samples_ran = sum of by_status over every status except 'unverified' (ran_draft_wrong and ran_fixture are failures of Syntology's instrument, not of the code); n_papers_ran = papers with at least one such sample = n_papers_ran_no_instrument_failure (papers with a run with no instrument failure) + n_papers_ran_instrument_only (papers where every run was a failure of Syntology's instrument); n_samples_ran = n_samples_ran_checked (honoured, violated, no contract) + n_samples_instrument (failures of Syntology's instrument, not of the code).","status_vocabulary":{"ran_honours":"ran, honoured the contract we drafted","ran_violates":"ran, violated the contract we drafted","ran_draft_wrong":"ran; our contract draft was wrong, not the code","ran_fixture":"ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)","ran":"ran on a synthesized input","unverified":"unverified (harvested, no recorded run)"},"n_papers":15,"n_papers_ran":14,"n_papers_ran_no_instrument_failure":7,"n_papers_ran_instrument_only":7,"units":"n_samples, n_samples_ran, n_samples_fingerprinted and by_status count distinct code bodies (code_sha256); n_places and n_places_pointer_only count places, one per (paper, code body) pair, which is also the unit of the samples list","n_samples":10,"n_samples_ran":9,"n_samples_constructed":0,"n_samples_ran_checked":5,"n_samples_instrument":4,"n_samples_unverified":1,"n_samples_fingerprinted":0,"n_places":15,"n_places_pointer_only":6,"n_places_pointer_only_link_open":2,"inline_ok_is":"inline_ok is true only when both licence gates allow Syntology to serve this code's text: this paper's link (inline_ok_link, from the licence of this paper's copy) and the stored copy of the same code that Syntology serves, which may have been first harvested from another repository. withheld_clause is Syntology's own wording of which gate keeps the text a pointer and why, and is null when inline_ok is true. This site shows no code text either way; n_places_pointer_only_link_open counts the places where this paper's link allows the text and the stored copy does not","by_status":{"ran_honours":0,"ran_violates":0,"ran_draft_wrong":4,"ran_fixture":0,"ran":5,"unverified":1},"key_notes":{"n_samples_ran_checked":"legacy name, kept unchanged so existing readers do not break: it counts the samples that ran with no instrument failure (honoured, violated, and ran with no contract checked); it does not mean a contract was checked, and the pages print it as 'K with no instrument failure', not 'K checked'","n_samples_constructed":"a sub-count of the samples that ran, never subtracted from them and never a failure: an executed sample whose run returned an instance of its own class (fixture_out_type equals the entry name): the run built an object and did not compute a result (Syntology's RAN record, counts.constructed)"},"syntology":{"atlas_url":null,"mcp":null,"mcp_per_sample":{"tool":"get_code","arguments_in":"samples[].mcp_get_code"},"developers":"https://syntology.ai/developers"},"samples":[{"arxiv_id":"2606.00535","arxiv_id_is_arxiv":true,"paper":"/paper/arxiv-2606-00535","title":"DREAM-S: Speculative Decoding with Searchable Drafting and Target-Aware Refinement for Multimodal Generation","date":null,"month_inferred_from_arxiv_id":"2026-06","title_source":"syntology","repo":"SAI-Lab-NYU/DREAM-S","path":"dream_s/model/kv_cache.py","file_url":"https://github.com/SAI-Lab-NYU/DREAM-S/blob/HEAD/dream_s/model/kv_cache.py","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"inline_ok_link":false,"withheld_clause":"GitHub detected no licence in this paper's repository, SAI-Lab-NYU/DREAM-S when it was checked","code_sha256":"2f5e7feccee982b7","code_sha256_prefix":"2f5e7feccee982b7","mcp_get_code":{"code_sha256":"2f5e7feccee982b7"}},{"arxiv_id":"2602.16052","arxiv_id_is_arxiv":true,"paper":"/paper/arxiv-2602-16052","title":"MoE-Spec: Expert Budgeting for Efficient Speculative Decoding","date":null,"month_inferred_from_arxiv_id":"2026-02","title_source":"syntology","repo":"SafeAILab/EAGLE","path":"eagle/model/kv_cache.py","file_url":"https://github.com/SafeAILab/EAGLE/blob/HEAD/eagle/model/kv_cache.py","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"inline_ok_link":false,"withheld_clause":"GitHub could not classify the licence of this paper's repository, SafeAILab/EAGLE (NOASSERTION)","code_sha256":"bbfbabefe7ad3955","code_sha256_prefix":"bbfbabefe7ad3955","mcp_get_code":{"code_sha256":"bbfbabefe7ad3955"}},{"arxiv_id":"2601.19278","arxiv_id_is_arxiv":true,"paper":"/paper/arxiv-2601-19278","title":"DART: Diffusion-Inspired Speculative Decoding for Fast LLM Inference","date":null,"month_inferred_from_arxiv_id":"2026-01","title_source":"syntology","repo":"fvliang/DART","path":"dart/model/kv_cache.py","file_url":"https://github.com/fvliang/DART/blob/HEAD/dart/model/kv_cache.py","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"inline_ok_link":true,"withheld_clause":null,"code_sha256":"9315d4be39a16b19","code_sha256_prefix":"9315d4be39a16b19","mcp_get_code":{"code_sha256":"9315d4be39a16b19"}},{"arxiv_id":"2601.15498","arxiv_id_is_arxiv":true,"paper":"/paper/arxiv-2601-15498","title":"MARS: Unleashing the Power of Speculative Decoding via Margin-Aware Verification","date":null,"month_inferred_from_arxiv_id":"2026-01","title_source":"syntology","repo":"5SSjw/MARS","path":"mars/model/kv_cache.py","file_url":"https://github.com/5SSjw/MARS/blob/HEAD/mars/model/kv_cache.py","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":false,"inline_ok_link":true,"withheld_clause":"this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from SafeAILab/EAGLE, and GitHub could not classify the licence of that repository (NOASSERTION)","code_sha256":"bbfbabefe7ad3955","code_sha256_prefix":"bbfbabefe7ad3955","mcp_get_code":{"code_sha256":"bbfbabefe7ad3955"}},{"arxiv_id":"2509.15235","arxiv_id_is_arxiv":true,"paper":"/paper/arxiv-2509-15235","title":"ViSpec: Accelerating Vision-Language Models with Vision-Aware Speculative Decoding","date":null,"month_inferred_from_arxiv_id":"2025-09","title_source":"syntology","repo":"KangJialiang/ViSpec","path":"vispec/model/kv_cache.py","file_url":"https://github.com/KangJialiang/ViSpec/blob/HEAD/vispec/model/kv_cache.py","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"inline_ok_link":true,"withheld_clause":null,"code_sha256":"86a9a36211cc9bb5","code_sha256_prefix":"86a9a36211cc9bb5","mcp_get_code":{"code_sha256":"86a9a36211cc9bb5"}},{"arxiv_id":"2508.16201","arxiv_id_is_arxiv":true,"paper":"/paper/arxiv-2508-16201","title":"SPECVLM: Enhancing Speculative Decoding of Video LLMs via Verifier-Guided Token Pruning","date":null,"month_inferred_from_arxiv_id":"2025-08","title_source":"syntology","repo":"zju-jiyicheng/SpecVLM","path":"decoding/tree_decoding.py","file_url":"https://github.com/zju-jiyicheng/SpecVLM/blob/HEAD/decoding/tree_decoding.py","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"inline_ok_link":false,"withheld_clause":"GitHub detected no licence in this paper's repository, zju-jiyicheng/SpecVLM when it was checked","code_sha256":"d992d12cd63a20e0","code_sha256_prefix":"d992d12cd63a20e0","mcp_get_code":{"code_sha256":"d992d12cd63a20e0"}},{"arxiv_id":"2507.22424","arxiv_id_is_arxiv":true,"paper":null,"title":"arXiv:2507.22424","date":null,"month_inferred_from_arxiv_id":"2025-07","title_source":null,"repo":"PineTreeWss/SpecVLA","path":"openvla/specdecoding/model/kv_cache.py","file_url":"https://github.com/PineTreeWss/SpecVLA/blob/HEAD/openvla/specdecoding/model/kv_cache.py","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"inline_ok_link":true,"withheld_clause":null,"code_sha256":"2fa0d32bf1353ed5","code_sha256_prefix":"2fa0d32bf1353ed5","mcp_get_code":{"code_sha256":"2fa0d32bf1353ed5"}},{"arxiv_id":"2503.10135","arxiv_id_is_arxiv":true,"paper":"/paper/gumiho-a-hybrid-architecture-to-prioritize","title":"Gumiho: A Hybrid Architecture to Prioritize Early Tokens in Speculative Decoding","date":"2025-03-13","month_inferred_from_arxiv_id":null,"title_source":"archive","repo":"AMD-AIG-AIMA/Gumiho","path":"gumiho/model/kv_cache.py","file_url":"https://github.com/AMD-AIG-AIMA/Gumiho/blob/HEAD/gumiho/model/kv_cache.py","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"inline_ok_link":true,"withheld_clause":null,"code_sha256":"aadc9c83d7ec2701","code_sha256_prefix":"aadc9c83d7ec2701","mcp_get_code":{"code_sha256":"aadc9c83d7ec2701"}},{"arxiv_id":"2410.06916","arxiv_id_is_arxiv":true,"paper":"/paper/swift-on-the-fly-self-speculative-decoding","title":"SWIFT: On-the-Fly Self-Speculative Decoding for LLM Inference Acceleration","date":"2024-10-09","month_inferred_from_arxiv_id":null,"title_source":"archive","repo":"hemingkx/SWIFT","path":"evaluation_llama/inference_swift.py","file_url":"https://github.com/hemingkx/SWIFT/blob/HEAD/evaluation_llama/inference_swift.py","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"inline_ok_link":true,"withheld_clause":null,"code_sha256":"aadc9c83d7ec2701","code_sha256_prefix":"aadc9c83d7ec2701","mcp_get_code":{"code_sha256":"aadc9c83d7ec2701"}},{"arxiv_id":"2408.00264","arxiv_id_is_arxiv":true,"paper":"/paper/clover-2-accurate-inference-for-regressive","title":"Clover-2: Accurate Inference for Regressive Lightweight Speculative Decoding","date":"2024-08-01","month_inferred_from_arxiv_id":null,"title_source":"archive","repo":"XiaoBin1992/clover","path":"clover/model/kv_cache.py","file_url":"https://github.com/XiaoBin1992/clover/blob/HEAD/clover/model/kv_cache.py","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"inline_ok_link":true,"withheld_clause":null,"code_sha256":"aadc9c83d7ec2701","code_sha256_prefix":"aadc9c83d7ec2701","mcp_get_code":{"code_sha256":"aadc9c83d7ec2701"}},{"arxiv_id":"2406.17404","arxiv_id_is_arxiv":true,"paper":"/paper/make-some-noise-unlocking-language-model","title":"Make Some Noise: Unlocking Language Model Parallel Inference Capability through Noisy Training","date":"2024-06-25","month_inferred_from_arxiv_id":null,"title_source":"archive","repo":"wyxstriker/MakeSomeNoiseInference","path":"model/noise_forward.py","file_url":"https://github.com/wyxstriker/MakeSomeNoiseInference/blob/HEAD/model/noise_forward.py","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"inline_ok_link":false,"withheld_clause":"GitHub detected no licence in this paper's repository, wyxstriker/MakeSomeNoiseInference when it was checked","code_sha256":"56a9074f8ed59e83","code_sha256_prefix":"56a9074f8ed59e83","mcp_get_code":{"code_sha256":"56a9074f8ed59e83"}},{"arxiv_id":"2402.05109","arxiv_id_is_arxiv":true,"paper":"/paper/hydra-sequentially-dependent-draft-heads-for","title":"Hydra: Sequentially-Dependent Draft Heads for Medusa Decoding","date":"2024-02-07","month_inferred_from_arxiv_id":null,"title_source":"archive","repo":"zankner/hydra","path":"hydra/model/kv_cache.py","file_url":"https://github.com/zankner/hydra/blob/HEAD/hydra/model/kv_cache.py","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"inline_ok_link":true,"withheld_clause":null,"code_sha256":"d2d232761fe51ccb","code_sha256_prefix":"d2d232761fe51ccb","mcp_get_code":{"code_sha256":"d2d232761fe51ccb"}},{"arxiv_id":"2401.10774","arxiv_id_is_arxiv":true,"paper":"/paper/medusa-simple-llm-inference-acceleration","title":"Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads","date":"2024-01-19","month_inferred_from_arxiv_id":null,"title_source":"archive","repo":"fasterdecoding/medusa","path":"medusa/model/kv_cache.py","file_url":"https://github.com/fasterdecoding/medusa/blob/HEAD/medusa/model/kv_cache.py","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":false,"inline_ok_link":true,"withheld_clause":"this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from wyxstriker/MakeSomeNoiseInference, and GitHub detected no licence in that repository when it was checked","code_sha256":"56a9074f8ed59e83","code_sha256_prefix":"56a9074f8ed59e83","mcp_get_code":{"code_sha256":"56a9074f8ed59e83"}},{"arxiv_id":"2311.08252","arxiv_id_is_arxiv":true,"paper":"/paper/rest-retrieval-based-speculative-decoding","title":"REST: Retrieval-Based Speculative Decoding","date":"2023-11-14","month_inferred_from_arxiv_id":null,"title_source":"archive","repo":"fasterdecoding/rest","path":"rest/model/kv_cache.py","file_url":"https://github.com/fasterdecoding/rest/blob/HEAD/rest/model/kv_cache.py","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"inline_ok_link":true,"withheld_clause":null,"code_sha256":"3965aca677b30513","code_sha256_prefix":"3965aca677b30513","mcp_get_code":{"code_sha256":"3965aca677b30513"}},{"arxiv_id":"2024.naacl-long.88","arxiv_id_is_arxiv":false,"paper":null,"title":"ACL Anthology id 2024.naacl-long.88","date":null,"month_inferred_from_arxiv_id":null,"title_source":null,"repo":"FasterDecoding/REST","path":"rest/model/kv_cache.py","file_url":"https://github.com/FasterDecoding/REST/blob/HEAD/rest/model/kv_cache.py","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"inline_ok_link":true,"withheld_clause":null,"code_sha256":"3965aca677b30513","code_sha256_prefix":"3965aca677b30513","mcp_get_code":{"code_sha256":"3965aca677b30513"}}],"record_sha256":"cc5febae02d54f0bd7d921ff46fdaaa1d5a51fde0d9c65438a2c7426a0d9f3f6","record_changed_at":"2026-10-05","record_changed_at_basis":"changed"}