{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:A7V2254YD5WHMLNE5DFW5Z5X37","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"3f929fedd8949b62128e5a9887c5e2203e6e36e004f53e0ef44725a255407b01","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-16T03:43:10Z","title_canon_sha256":"e169d93e2760b75e10d1dd3c072fb954d0b34eeb2ef0500afd4663bbc98af489"},"schema_version":"1.0","source":{"id":"2605.16786","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.16786","created_at":"2026-05-20T00:03:21Z"},{"alias_kind":"arxiv_version","alias_value":"2605.16786v1","created_at":"2026-05-20T00:03:21Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.16786","created_at":"2026-05-20T00:03:21Z"},{"alias_kind":"pith_short_12","alias_value":"A7V2254YD5WH","created_at":"2026-05-20T00:03:21Z"},{"alias_kind":"pith_short_16","alias_value":"A7V2254YD5WHMLNE","created_at":"2026-05-20T00:03:21Z"},{"alias_kind":"pith_short_8","alias_value":"A7V2254Y","created_at":"2026-05-20T00:03:21Z"}],"graph_snapshots":[{"event_id":"sha256:a56d46888e772cc6ffca8472c9e498b1af850f0e693bb459bc187d41e62650c8","target":"graph","created_at":"2026-05-20T00:03:21Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Lever reduces inference latency by an average of 2.93x over baseline flash-offloaded inference and 1.50x over conventional speculative decoding, narrowing the latency gap between flash-backed and memory-resident LLM inference."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The assumption that jointly optimizing token-tree construction with an I/O- and compute-aware gain-cost objective, early-exit pruning, and CPU-NPU mapping will deliver the claimed speedups under real smartphone I/O latency and parallelism constraints (stated in the abstract description of the three stages)."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Lever optimizes the drafting, verification, and execution stages of speculative decoding for flash-backed LLM inference on smartphones, reporting 2.93x average latency reduction over baseline flash-offloaded inference."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Lever reduces smartphone LLM inference latency by 2.93x over flash baselines through optimized speculative decoding."}],"snapshot_sha256":"22321fe8fb1e2e705f5dfa4b3ccc5f2c8453f4102283486d727b4110e21921db"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"fd5d216180b46a71e3f4c9ed90582bafdf96906dc54352ca62d6495bfa29330d"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T22:01:19.651802Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T21:50:55.144814Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-19T19:01:56.298520Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T18:33:26.434219Z","status":"skipped","version":"1.0.0"}],"endpoint":"/pith/2605.16786/integrity.json","findings":[],"snapshot_sha256":"033653024c54694c2ab0b8aaa7f00a7e56e2289d5ac785aec6ad7a1a8e2a07fb","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Large language models (LLMs) are increasingly needed for interactive mobile applications, but high-quality models exceed the limited DRAM available on smartphones. Flash storage can hold larger models, yet flash-backed inference is slow because autoregressive decoding repeatedly invokes the target model and incurs costly I/O. We observe that speculative decoding is a natural fit for this setting: a small draft model can remain in DRAM, while a larger flash-resident target model verifies multiple candidate tokens per invocation. However, existing methods assume server-class accelerators and fai","authors_text":"Fengzu Li, Ju Ren, Tuowei Wang, Wei Gao, Yanfan Sun","cross_cats":[],"headline":"Lever reduces smartphone LLM inference latency by 2.93x over flash baselines through optimized speculative decoding.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-16T03:43:10Z","title":"Lever: Speculative LLM Inference on Smartphones"},"references":{"count":40,"internal_anchors":8,"resolved_work":40,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Llm in a flash: Efficient large language model inference with limited memory","work_id":"a9c527dd-dab9-4723-894f-c136b3e1ccaf","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Hydra: Sequentially-dependent draft heads for medusa decoding","work_id":"c49cff9c-dcd0-4434-ad6b-51534aca2b2f","year":2024},{"cited_arxiv_id":"2108.07732","doi":"","is_internal_anchor":true,"ref_index":3,"title":"Program Synthesis with Large Language Models","work_id":"fd241a05-03b9-4de2-9588-9d77ce176125","year":2021},{"cited_arxiv_id":"2401.10774","doi":"","is_internal_anchor":true,"ref_index":4,"title":"Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads","work_id":"f6202cd1-1a78-4c19-8242-688acf4952b6","year":2024},{"cited_arxiv_id":"2302.01318","doi":"","is_internal_anchor":true,"ref_index":5,"title":"Accelerating Large Language Model Decoding with Speculative Sampling","work_id":"b53b11e5-8613-439d-9d9d-0e2a9090d79f","year":2023}],"snapshot_sha256":"cf7636b1bc1a7d1a0adbce931c825d59311bc69e86f4a77b84ddec39c92c4cee"},"source":{"id":"2605.16786","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-19T21:39:51.959093Z","id":"aa5ea424-dbb2-41ce-b6a5-16fd26eabf79","model_set":{"reader":"grok-4.3"},"one_line_summary":"Lever optimizes the drafting, verification, and execution stages of speculative decoding for flash-backed LLM inference on smartphones, reporting 2.93x average latency reduction over baseline flash-offloaded inference.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Lever reduces smartphone LLM inference latency by 2.93x over flash baselines through optimized speculative decoding.","strongest_claim":"Lever reduces inference latency by an average of 2.93x over baseline flash-offloaded inference and 1.50x over conventional speculative decoding, narrowing the latency gap between flash-backed and memory-resident LLM inference.","weakest_assumption":"The assumption that jointly optimizing token-tree construction with an I/O- and compute-aware gain-cost objective, early-exit pruning, and CPU-NPU mapping will deliver the claimed speedups under real smartphone I/O latency and parallelism constraints (stated in the abstract description of the three stages)."}},"verdict_id":"aa5ea424-dbb2-41ce-b6a5-16fd26eabf79"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:9c6323176f50ae35daa3de7f84388362ea6803e974e018a23541e948b5269273","target":"record","created_at":"2026-05-20T00:03:21Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"3f929fedd8949b62128e5a9887c5e2203e6e36e004f53e0ef44725a255407b01","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-16T03:43:10Z","title_canon_sha256":"e169d93e2760b75e10d1dd3c072fb954d0b34eeb2ef0500afd4663bbc98af489"},"schema_version":"1.0","source":{"id":"2605.16786","kind":"arxiv","version":1}},"canonical_sha256":"07ebad77981f6c762da4e8cb6ee7b7dfeb5aa04bbd6b90890cba5937f281cb64","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"07ebad77981f6c762da4e8cb6ee7b7dfeb5aa04bbd6b90890cba5937f281cb64","first_computed_at":"2026-05-20T00:03:21.936503Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:03:21.936503Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"7/iSMpI0rDKIHH/W0V7k40cA6z/lsd+QW1upW0oP4cdtbtQKq5ikP0vdzNUTbhNwke/HHPc4IT8t51LPN5ZSBg==","signature_status":"signed_v1","signed_at":"2026-05-20T00:03:21.937389Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.16786","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:9c6323176f50ae35daa3de7f84388362ea6803e974e018a23541e948b5269273","sha256:a56d46888e772cc6ffca8472c9e498b1af850f0e693bb459bc187d41e62650c8"],"state_sha256":"4ea1ba6c75312a9fc1f240c3123b264a0e3db5c8e96217c6e4f01f6c57f48534"}