{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2021:AGKTK4CAEUVVVSDCEM4WDIY5EA","short_pith_number":"pith:AGKTK4CA","canonical_record":{"source":{"id":"2111.00230","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-10-30T11:07:43Z","cross_cats_sorted":[],"title_canon_sha256":"69cf05d9732db1497fff84d43b68efa9aa6aa8cc36cd42d1d34ae679aa9efc98","abstract_canon_sha256":"9bbd6cac73c111694d46cc28280530b1b5948d31e20d40b729345c0c69f652e0"},"schema_version":"1.0"},"canonical_sha256":"0195357040252b5ac862233961a31d200cac91ab7fb6c413a9d15e2eddbe7924","source":{"kind":"arxiv","id":"2111.00230","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2111.00230","created_at":"2026-07-05T03:27:39Z"},{"alias_kind":"arxiv_version","alias_value":"2111.00230v1","created_at":"2026-07-05T03:27:39Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2111.00230","created_at":"2026-07-05T03:27:39Z"},{"alias_kind":"pith_short_12","alias_value":"AGKTK4CAEUVV","created_at":"2026-07-05T03:27:39Z"},{"alias_kind":"pith_short_16","alias_value":"AGKTK4CAEUVVVSDC","created_at":"2026-07-05T03:27:39Z"},{"alias_kind":"pith_short_8","alias_value":"AGKTK4CA","created_at":"2026-07-05T03:27:39Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2021:AGKTK4CAEUVVVSDCEM4WDIY5EA","target":"record","payload":{"canonical_record":{"source":{"id":"2111.00230","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-10-30T11:07:43Z","cross_cats_sorted":[],"title_canon_sha256":"69cf05d9732db1497fff84d43b68efa9aa6aa8cc36cd42d1d34ae679aa9efc98","abstract_canon_sha256":"9bbd6cac73c111694d46cc28280530b1b5948d31e20d40b729345c0c69f652e0"},"schema_version":"1.0"},"canonical_sha256":"0195357040252b5ac862233961a31d200cac91ab7fb6c413a9d15e2eddbe7924","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:27:39.816630Z","signature_b64":"jHmd4ciw8FgdCEnnqL2Zn3dq9temfoTPhLozIs0ibA/oVGdmIWRSHJ3XRpYL9auO7Rp2xfJy8q0406xuHBzlBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0195357040252b5ac862233961a31d200cac91ab7fb6c413a9d15e2eddbe7924","last_reissued_at":"2026-07-05T03:27:39.815872Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:27:39.815872Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2111.00230","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T03:27:39Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"WdOHByouWGch70yojVta3kHEjyAQbW6jPNjSurRsBdh5X4BCjwomjsZiE8hrffNLTRCED5tlcTnJmXJB64KKAg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-07T07:16:34.255242Z"},"content_sha256":"e9b07b7074e55e97b38f103fd39998e6f8ffe736d5e512286433356877b6872b","schema_version":"1.0","event_id":"sha256:e9b07b7074e55e97b38f103fd39998e6f8ffe736d5e512286433356877b6872b"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2021:AGKTK4CAEUVVVSDCEM4WDIY5EA","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Magic Pyramid: Accelerating Inference with Early Exiting and Token Pruning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Belinda Zeng, Iman Keivanloo, Santosh Rajagopalan, Trishul Chilimbi, Xiang He, Xuanli He, Yi Xu","submitted_at":"2021-10-30T11:07:43Z","abstract_excerpt":"Pre-training and then fine-tuning large language models is commonly used to achieve state-of-the-art performance in natural language processing (NLP) tasks. However, most pre-trained models suffer from low inference speed. Deploying such large models to applications with latency constraints is challenging. In this work, we focus on accelerating the inference via conditional computations. To achieve this, we propose a novel idea, Magic Pyramid (MP), to reduce both width-wise and depth-wise computation via token pruning and early exiting for Transformer-based models, particularly BERT. The forme"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2111.00230","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/2111.00230/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T03:27:39Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"dtilsPqKLYRk+QaagsshHR1GYAAbAtTZKeSnyBcm8q/miWpkNmbdKhdC9aJUW2JVr/wtPbGhB0SNpt4RU4TqAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-07T07:16:34.255905Z"},"content_sha256":"2e77d0eea7e382e7ece41cd53d97a6286538b76e4ff1d30a2b3054c1c5ac3a22","schema_version":"1.0","event_id":"sha256:2e77d0eea7e382e7ece41cd53d97a6286538b76e4ff1d30a2b3054c1c5ac3a22"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/AGKTK4CAEUVVVSDCEM4WDIY5EA/bundle.json","state_url":"https://pith.science/pith/AGKTK4CAEUVVVSDCEM4WDIY5EA/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/AGKTK4CAEUVVVSDCEM4WDIY5EA/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-07-07T07:16:34Z","links":{"resolver":"https://pith.science/pith/AGKTK4CAEUVVVSDCEM4WDIY5EA","bundle":"https://pith.science/pith/AGKTK4CAEUVVVSDCEM4WDIY5EA/bundle.json","state":"https://pith.science/pith/AGKTK4CAEUVVVSDCEM4WDIY5EA/state.json","well_known_bundle":"https://pith.science/.well-known/pith/AGKTK4CAEUVVVSDCEM4WDIY5EA/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2021:AGKTK4CAEUVVVSDCEM4WDIY5EA","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":"9bbd6cac73c111694d46cc28280530b1b5948d31e20d40b729345c0c69f652e0","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-10-30T11:07:43Z","title_canon_sha256":"69cf05d9732db1497fff84d43b68efa9aa6aa8cc36cd42d1d34ae679aa9efc98"},"schema_version":"1.0","source":{"id":"2111.00230","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2111.00230","created_at":"2026-07-05T03:27:39Z"},{"alias_kind":"arxiv_version","alias_value":"2111.00230v1","created_at":"2026-07-05T03:27:39Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2111.00230","created_at":"2026-07-05T03:27:39Z"},{"alias_kind":"pith_short_12","alias_value":"AGKTK4CAEUVV","created_at":"2026-07-05T03:27:39Z"},{"alias_kind":"pith_short_16","alias_value":"AGKTK4CAEUVVVSDC","created_at":"2026-07-05T03:27:39Z"},{"alias_kind":"pith_short_8","alias_value":"AGKTK4CA","created_at":"2026-07-05T03:27:39Z"}],"graph_snapshots":[{"event_id":"sha256:2e77d0eea7e382e7ece41cd53d97a6286538b76e4ff1d30a2b3054c1c5ac3a22","target":"graph","created_at":"2026-07-05T03:27:39Z","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":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2111.00230/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Pre-training and then fine-tuning large language models is commonly used to achieve state-of-the-art performance in natural language processing (NLP) tasks. However, most pre-trained models suffer from low inference speed. Deploying such large models to applications with latency constraints is challenging. In this work, we focus on accelerating the inference via conditional computations. To achieve this, we propose a novel idea, Magic Pyramid (MP), to reduce both width-wise and depth-wise computation via token pruning and early exiting for Transformer-based models, particularly BERT. The forme","authors_text":"Belinda Zeng, Iman Keivanloo, Santosh Rajagopalan, Trishul Chilimbi, Xiang He, Xuanli He, Yi Xu","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-10-30T11:07:43Z","title":"Magic Pyramid: Accelerating Inference with Early Exiting and Token Pruning"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2111.00230","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:e9b07b7074e55e97b38f103fd39998e6f8ffe736d5e512286433356877b6872b","target":"record","created_at":"2026-07-05T03:27:39Z","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":"9bbd6cac73c111694d46cc28280530b1b5948d31e20d40b729345c0c69f652e0","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-10-30T11:07:43Z","title_canon_sha256":"69cf05d9732db1497fff84d43b68efa9aa6aa8cc36cd42d1d34ae679aa9efc98"},"schema_version":"1.0","source":{"id":"2111.00230","kind":"arxiv","version":1}},"canonical_sha256":"0195357040252b5ac862233961a31d200cac91ab7fb6c413a9d15e2eddbe7924","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"0195357040252b5ac862233961a31d200cac91ab7fb6c413a9d15e2eddbe7924","first_computed_at":"2026-07-05T03:27:39.815872Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T03:27:39.815872Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"jHmd4ciw8FgdCEnnqL2Zn3dq9temfoTPhLozIs0ibA/oVGdmIWRSHJ3XRpYL9auO7Rp2xfJy8q0406xuHBzlBA==","signature_status":"signed_v1","signed_at":"2026-07-05T03:27:39.816630Z","signed_message":"canonical_sha256_bytes"},"source_id":"2111.00230","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:e9b07b7074e55e97b38f103fd39998e6f8ffe736d5e512286433356877b6872b","sha256:2e77d0eea7e382e7ece41cd53d97a6286538b76e4ff1d30a2b3054c1c5ac3a22"],"state_sha256":"b91601a1e0a046ec67af556c1f6f587bdfdf023fa45288272cc7dd302c5d4d24"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"YiF1aIYG0YpzAf5l8kbMXxSwbrhQpUQrFGDflBGAlw9sqqyivxvx/zYRixzhuDPz6nENzcygl/akLl5MOBM3BQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-07T07:16:34.259612Z","bundle_sha256":"e81bd29ca8ad74412a56a95e53988be84d14a974b4ff01906579362773109240"}}