{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:YN37UFVISQLXYGNTVVIXS6PGOO","short_pith_number":"pith:YN37UFVI","schema_version":"1.0","canonical_sha256":"c377fa16a894177c19b3ad517979e67394ef670bcc09f940625dd852e19e2a44","source":{"kind":"arxiv","id":"2404.10136","version":1},"attestation_state":"computed","paper":{"title":"Language Model Cascades: Token-level uncertainty and beyond","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Aditya Krishna Menon, Ankit Singh Rawat, Harikrishna Narasimhan, Neha Gupta, Sanjiv Kumar, Wittawat Jitkrittum","submitted_at":"2024-04-15T21:02:48Z","abstract_excerpt":"Recent advances in language models (LMs) have led to significant improvements in quality on complex NLP tasks, but at the expense of increased inference costs. Cascading offers a simple strategy to achieve more favorable cost-quality tradeoffs: here, a small model is invoked for most \"easy\" instances, while a few \"hard\" instances are deferred to the large model. While the principles underpinning cascading are well-studied for classification tasks - with deferral based on predicted class uncertainty favored theoretically and practically - a similar understanding is lacking for generative LM tas"},"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":"2404.10136","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-04-15T21:02:48Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"793f4e08b5050041b5b9233c79d619c10d4ff3b1d8170f50033b41bb72dcab30","abstract_canon_sha256":"a75d6fc00ba08b3f8afc14e2b07868e7bcfbdb23eb6c60d88e817300687eedbb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:08:25.913835Z","signature_b64":"mKAMumKR63LguZhZ8pz695JSj2p5FtUblOMu4VUaSGYoLFPFrhGykwfDd/5nVVyRZO8bllP6AQ7PZBxCkwtaAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c377fa16a894177c19b3ad517979e67394ef670bcc09f940625dd852e19e2a44","last_reissued_at":"2026-07-05T08:08:25.910891Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:08:25.910891Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Language Model Cascades: Token-level uncertainty and beyond","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Aditya Krishna Menon, Ankit Singh Rawat, Harikrishna Narasimhan, Neha Gupta, Sanjiv Kumar, Wittawat Jitkrittum","submitted_at":"2024-04-15T21:02:48Z","abstract_excerpt":"Recent advances in language models (LMs) have led to significant improvements in quality on complex NLP tasks, but at the expense of increased inference costs. Cascading offers a simple strategy to achieve more favorable cost-quality tradeoffs: here, a small model is invoked for most \"easy\" instances, while a few \"hard\" instances are deferred to the large model. While the principles underpinning cascading are well-studied for classification tasks - with deferral based on predicted class uncertainty favored theoretically and practically - a similar understanding is lacking for generative LM tas"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.10136","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/2404.10136/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":"2404.10136","created_at":"2026-07-05T08:08:25.913335+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.10136v1","created_at":"2026-07-05T08:08:25.913335+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.10136","created_at":"2026-07-05T08:08:25.913335+00:00"},{"alias_kind":"pith_short_12","alias_value":"YN37UFVISQLX","created_at":"2026-07-05T08:08:25.913335+00:00"},{"alias_kind":"pith_short_16","alias_value":"YN37UFVISQLXYGNT","created_at":"2026-07-05T08:08:25.913335+00:00"},{"alias_kind":"pith_short_8","alias_value":"YN37UFVI","created_at":"2026-07-05T08:08:25.913335+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.12243","citing_title":"VIA-SD: Verification via Intra-Model Routing for Speculative Decoding","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07392","citing_title":"Online Pandora's Box for Contextual LLM Cascading","ref_index":115,"is_internal_anchor":false},{"citing_arxiv_id":"2502.18036","citing_title":"Harnessing Multiple Large Language Models: A Survey on LLM Ensemble","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17288","citing_title":"When Efficiency Backfires: Cascading LLMs Trigger Cascade Failure under Adversarial Attack","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2506.13351","citing_title":"Direct Reasoning Optimization: Token-Level Reasoning Reflectivity Meets Rubric Gates for Unverifiable Tasks","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05777","citing_title":"Estimating the Black-box LLM Uncertainty with Distribution-Aligned Adversarial Distillation","ref_index":15,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YN37UFVISQLXYGNTVVIXS6PGOO","json":"https://pith.science/pith/YN37UFVISQLXYGNTVVIXS6PGOO.json","graph_json":"https://pith.science/api/pith-number/YN37UFVISQLXYGNTVVIXS6PGOO/graph.json","events_json":"https://pith.science/api/pith-number/YN37UFVISQLXYGNTVVIXS6PGOO/events.json","paper":"https://pith.science/paper/YN37UFVI"},"agent_actions":{"view_html":"https://pith.science/pith/YN37UFVISQLXYGNTVVIXS6PGOO","download_json":"https://pith.science/pith/YN37UFVISQLXYGNTVVIXS6PGOO.json","view_paper":"https://pith.science/paper/YN37UFVI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.10136&json=true","fetch_graph":"https://pith.science/api/pith-number/YN37UFVISQLXYGNTVVIXS6PGOO/graph.json","fetch_events":"https://pith.science/api/pith-number/YN37UFVISQLXYGNTVVIXS6PGOO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YN37UFVISQLXYGNTVVIXS6PGOO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YN37UFVISQLXYGNTVVIXS6PGOO/action/storage_attestation","attest_author":"https://pith.science/pith/YN37UFVISQLXYGNTVVIXS6PGOO/action/author_attestation","sign_citation":"https://pith.science/pith/YN37UFVISQLXYGNTVVIXS6PGOO/action/citation_signature","submit_replication":"https://pith.science/pith/YN37UFVISQLXYGNTVVIXS6PGOO/action/replication_record"}},"created_at":"2026-07-05T08:08:25.913335+00:00","updated_at":"2026-07-05T08:08:25.913335+00:00"}