{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:TPYR3ZNQOKTLO2AXLRYZ3A6KZN","short_pith_number":"pith:TPYR3ZNQ","schema_version":"1.0","canonical_sha256":"9bf11de5b072a6b768175c719d83cacb7aeb7928ffe7ec5bddaa8be8e604c84b","source":{"kind":"arxiv","id":"2607.14107","version":1},"attestation_state":"computed","paper":{"title":"Polestar: Drift-Aware Cache Calibration and Token Commitment for Efficient Inference of Diffusion LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Akshat Ramachandran, Mingyu Lee, Souvik Kundu, Tushar Krishna","submitted_at":"2026-05-07T18:05:39Z","abstract_excerpt":"The inference efficiency of diffusion large language models (dLLMs) is constrained by two challenges: bidirectional attention precludes efficient KV-cache reuse, while increasing decoding parallelism with static confidence thresholds can compromise generation quality. We observe that both challenges arise from a shared phenomenon: as tokens are decoded, their contextual integration through bidirectional attention causes token representations to drift (evolve) across decoding steps. This insight motivates Polestar, a training-free inference framework that uses token representation drift as a un"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2607.14107","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-07T18:05:39Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"6c24f196bee4fc465c8b36b8574b93746098184e1f64fd34fdff0c64449584a0","abstract_canon_sha256":"3dad07babc058f36668c0f18218be27d7d6da0c42024a189547fe17788f7906b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-17T00:20:52.400339Z","signature_b64":"7n7279RBB5rS3cqL8cOa2KPG+OsiQHwNZUOEs/ufTk8toxZOGWYN+wmzXmRHdAd09hY2cOnWvtBAGvVbNGXCBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9bf11de5b072a6b768175c719d83cacb7aeb7928ffe7ec5bddaa8be8e604c84b","last_reissued_at":"2026-07-17T00:20:52.399511Z","signature_status":"signed_v1","first_computed_at":"2026-07-17T00:20:52.399511Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Polestar: Drift-Aware Cache Calibration and Token Commitment for Efficient Inference of Diffusion LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Akshat Ramachandran, Mingyu Lee, Souvik Kundu, Tushar Krishna","submitted_at":"2026-05-07T18:05:39Z","abstract_excerpt":"The inference efficiency of diffusion large language models (dLLMs) is constrained by two challenges: bidirectional attention precludes efficient KV-cache reuse, while increasing decoding parallelism with static confidence thresholds can compromise generation quality. We observe that both challenges arise from a shared phenomenon: as tokens are decoded, their contextual integration through bidirectional attention causes token representations to drift (evolve) across decoding steps. This insight motivates Polestar, a training-free inference framework that uses token representation drift as a un"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.14107","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2607.14107/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2607.14107","created_at":"2026-07-17T00:20:52.399926+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.14107v1","created_at":"2026-07-17T00:20:52.399926+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.14107","created_at":"2026-07-17T00:20:52.399926+00:00"},{"alias_kind":"pith_short_12","alias_value":"TPYR3ZNQOKTL","created_at":"2026-07-17T00:20:52.399926+00:00"},{"alias_kind":"pith_short_16","alias_value":"TPYR3ZNQOKTLO2AX","created_at":"2026-07-17T00:20:52.399926+00:00"},{"alias_kind":"pith_short_8","alias_value":"TPYR3ZNQ","created_at":"2026-07-17T00:20:52.399926+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TPYR3ZNQOKTLO2AXLRYZ3A6KZN","json":"https://pith.science/pith/TPYR3ZNQOKTLO2AXLRYZ3A6KZN.json","graph_json":"https://pith.science/api/pith-number/TPYR3ZNQOKTLO2AXLRYZ3A6KZN/graph.json","events_json":"https://pith.science/api/pith-number/TPYR3ZNQOKTLO2AXLRYZ3A6KZN/events.json","paper":"https://pith.science/paper/TPYR3ZNQ"},"agent_actions":{"view_html":"https://pith.science/pith/TPYR3ZNQOKTLO2AXLRYZ3A6KZN","download_json":"https://pith.science/pith/TPYR3ZNQOKTLO2AXLRYZ3A6KZN.json","view_paper":"https://pith.science/paper/TPYR3ZNQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.14107&json=true","fetch_graph":"https://pith.science/api/pith-number/TPYR3ZNQOKTLO2AXLRYZ3A6KZN/graph.json","fetch_events":"https://pith.science/api/pith-number/TPYR3ZNQOKTLO2AXLRYZ3A6KZN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TPYR3ZNQOKTLO2AXLRYZ3A6KZN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TPYR3ZNQOKTLO2AXLRYZ3A6KZN/action/storage_attestation","attest_author":"https://pith.science/pith/TPYR3ZNQOKTLO2AXLRYZ3A6KZN/action/author_attestation","sign_citation":"https://pith.science/pith/TPYR3ZNQOKTLO2AXLRYZ3A6KZN/action/citation_signature","submit_replication":"https://pith.science/pith/TPYR3ZNQOKTLO2AXLRYZ3A6KZN/action/replication_record"}},"created_at":"2026-07-17T00:20:52.399926+00:00","updated_at":"2026-07-17T00:20:52.399926+00:00"}