{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:KDUG44MTLTQLOAK6UXH5U5UNLG","short_pith_number":"pith:KDUG44MT","schema_version":"1.0","canonical_sha256":"50e86e71935ce0b7015ea5cfda768d59824199b9a8ba1fb7730d3a9cfed97d65","source":{"kind":"arxiv","id":"2502.02040","version":1},"attestation_state":"computed","paper":{"title":"M2R2: Mixture of Multi-Rate Residuals for Efficient Transformer Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Devang Naik, Irina Belousova, Mahyar Najibi, Nikhil Bhendawade","submitted_at":"2025-02-04T06:13:52Z","abstract_excerpt":"Residual transformations enhance the representational depth and expressive power of large language models (LLMs). However, applying static residual transformations across all tokens in auto-regressive generation leads to a suboptimal trade-off between inference efficiency and generation fidelity. Existing methods, including Early Exiting, Skip Decoding, and Mixture-of-Depth address this by modulating the residual transformation based on token-level complexity. Nevertheless, these approaches predominantly consider the distance traversed by tokens through the model layers, neglecting the underly"},"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":"2502.02040","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-02-04T06:13:52Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"2f9b2fa72d6b6254add972e0f35afa864b5370dd4d9d3466e365ff3818cd0756","abstract_canon_sha256":"72a3a77f2c61d39f23e1820d955c63cbf8bf704db45e2072b84cda2f79ffcabc"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:09:22.660817Z","signature_b64":"RDW68RCTNPdDR5wfiia8xRwNFwMloYhBrpBGMV+kqrgFxBJau+7ESZAY65Mb2EVYZsj58x3c5KFtmjzE2+exBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"50e86e71935ce0b7015ea5cfda768d59824199b9a8ba1fb7730d3a9cfed97d65","last_reissued_at":"2026-07-05T10:09:22.660295Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:09:22.660295Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"M2R2: Mixture of Multi-Rate Residuals for Efficient Transformer Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Devang Naik, Irina Belousova, Mahyar Najibi, Nikhil Bhendawade","submitted_at":"2025-02-04T06:13:52Z","abstract_excerpt":"Residual transformations enhance the representational depth and expressive power of large language models (LLMs). However, applying static residual transformations across all tokens in auto-regressive generation leads to a suboptimal trade-off between inference efficiency and generation fidelity. Existing methods, including Early Exiting, Skip Decoding, and Mixture-of-Depth address this by modulating the residual transformation based on token-level complexity. Nevertheless, these approaches predominantly consider the distance traversed by tokens through the model layers, neglecting the underly"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.02040","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/2502.02040/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":"2502.02040","created_at":"2026-07-05T10:09:22.660365+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.02040v1","created_at":"2026-07-05T10:09:22.660365+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.02040","created_at":"2026-07-05T10:09:22.660365+00:00"},{"alias_kind":"pith_short_12","alias_value":"KDUG44MTLTQL","created_at":"2026-07-05T10:09:22.660365+00:00"},{"alias_kind":"pith_short_16","alias_value":"KDUG44MTLTQLOAK6","created_at":"2026-07-05T10:09:22.660365+00:00"},{"alias_kind":"pith_short_8","alias_value":"KDUG44MT","created_at":"2026-07-05T10:09:22.660365+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/KDUG44MTLTQLOAK6UXH5U5UNLG","json":"https://pith.science/pith/KDUG44MTLTQLOAK6UXH5U5UNLG.json","graph_json":"https://pith.science/api/pith-number/KDUG44MTLTQLOAK6UXH5U5UNLG/graph.json","events_json":"https://pith.science/api/pith-number/KDUG44MTLTQLOAK6UXH5U5UNLG/events.json","paper":"https://pith.science/paper/KDUG44MT"},"agent_actions":{"view_html":"https://pith.science/pith/KDUG44MTLTQLOAK6UXH5U5UNLG","download_json":"https://pith.science/pith/KDUG44MTLTQLOAK6UXH5U5UNLG.json","view_paper":"https://pith.science/paper/KDUG44MT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.02040&json=true","fetch_graph":"https://pith.science/api/pith-number/KDUG44MTLTQLOAK6UXH5U5UNLG/graph.json","fetch_events":"https://pith.science/api/pith-number/KDUG44MTLTQLOAK6UXH5U5UNLG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KDUG44MTLTQLOAK6UXH5U5UNLG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KDUG44MTLTQLOAK6UXH5U5UNLG/action/storage_attestation","attest_author":"https://pith.science/pith/KDUG44MTLTQLOAK6UXH5U5UNLG/action/author_attestation","sign_citation":"https://pith.science/pith/KDUG44MTLTQLOAK6UXH5U5UNLG/action/citation_signature","submit_replication":"https://pith.science/pith/KDUG44MTLTQLOAK6UXH5U5UNLG/action/replication_record"}},"created_at":"2026-07-05T10:09:22.660365+00:00","updated_at":"2026-07-05T10:09:22.660365+00:00"}