{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:7FHFSF23I6ESFULGQAF7UQXU2F","short_pith_number":"pith:7FHFSF23","schema_version":"1.0","canonical_sha256":"f94e59175b478922d166800bfa42f4d17182dd6aeaf5975863a67daa73d6db21","source":{"kind":"arxiv","id":"2502.15618","version":1},"attestation_state":"computed","paper":{"title":"Probe Pruning: Accelerating LLMs through Dynamic Pruning via Model-Probing","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Ali Anwar, Enmao Diao, Jie Ding, Li Yang, Qi Le, Xinran Wang, Ziyan Wang","submitted_at":"2025-02-21T17:41:21Z","abstract_excerpt":"We introduce Probe Pruning (PP), a novel framework for online, dynamic, structured pruning of Large Language Models (LLMs) applied in a batch-wise manner. PP leverages the insight that not all samples and tokens contribute equally to the model's output, and probing a small portion of each batch effectively identifies crucial weights, enabling tailored dynamic pruning for different batches. It comprises three main stages: probing, history-informed pruning, and full inference. In the probing stage, PP selects a small yet crucial set of hidden states, based on residual importance, to run a few mo"},"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":"2502.15618","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-02-21T17:41:21Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"740fec6423444ee0d5714e7a9c1d311921cd28e7b773e6aaf10d42a0b4ae4728","abstract_canon_sha256":"de8b71a51769e2e7b9fd4ba86c8c3cb424b9aed01b7a1780c5e4e824e9079e88"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:18:04.757615Z","signature_b64":"PfkaKDrZ2+SHo8dGFHUWcPRGWP2ueubNPuXBAGfrpIRCGGi/ydptSz/qOU7JMOMeOa8BzrQ3augQDLvoTeBYAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f94e59175b478922d166800bfa42f4d17182dd6aeaf5975863a67daa73d6db21","last_reissued_at":"2026-07-05T10:18:04.757076Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:18:04.757076Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Probe Pruning: Accelerating LLMs through Dynamic Pruning via Model-Probing","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Ali Anwar, Enmao Diao, Jie Ding, Li Yang, Qi Le, Xinran Wang, Ziyan Wang","submitted_at":"2025-02-21T17:41:21Z","abstract_excerpt":"We introduce Probe Pruning (PP), a novel framework for online, dynamic, structured pruning of Large Language Models (LLMs) applied in a batch-wise manner. PP leverages the insight that not all samples and tokens contribute equally to the model's output, and probing a small portion of each batch effectively identifies crucial weights, enabling tailored dynamic pruning for different batches. It comprises three main stages: probing, history-informed pruning, and full inference. In the probing stage, PP selects a small yet crucial set of hidden states, based on residual importance, to run a few mo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.15618","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/2502.15618/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":"2502.15618","created_at":"2026-07-05T10:18:04.757133+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.15618v1","created_at":"2026-07-05T10:18:04.757133+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.15618","created_at":"2026-07-05T10:18:04.757133+00:00"},{"alias_kind":"pith_short_12","alias_value":"7FHFSF23I6ES","created_at":"2026-07-05T10:18:04.757133+00:00"},{"alias_kind":"pith_short_16","alias_value":"7FHFSF23I6ESFULG","created_at":"2026-07-05T10:18:04.757133+00:00"},{"alias_kind":"pith_short_8","alias_value":"7FHFSF23","created_at":"2026-07-05T10:18:04.757133+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2505.17138","citing_title":"RAP: Runtime Adaptive Pruning for LLM Inference","ref_index":18,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7FHFSF23I6ESFULGQAF7UQXU2F","json":"https://pith.science/pith/7FHFSF23I6ESFULGQAF7UQXU2F.json","graph_json":"https://pith.science/api/pith-number/7FHFSF23I6ESFULGQAF7UQXU2F/graph.json","events_json":"https://pith.science/api/pith-number/7FHFSF23I6ESFULGQAF7UQXU2F/events.json","paper":"https://pith.science/paper/7FHFSF23"},"agent_actions":{"view_html":"https://pith.science/pith/7FHFSF23I6ESFULGQAF7UQXU2F","download_json":"https://pith.science/pith/7FHFSF23I6ESFULGQAF7UQXU2F.json","view_paper":"https://pith.science/paper/7FHFSF23","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.15618&json=true","fetch_graph":"https://pith.science/api/pith-number/7FHFSF23I6ESFULGQAF7UQXU2F/graph.json","fetch_events":"https://pith.science/api/pith-number/7FHFSF23I6ESFULGQAF7UQXU2F/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7FHFSF23I6ESFULGQAF7UQXU2F/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7FHFSF23I6ESFULGQAF7UQXU2F/action/storage_attestation","attest_author":"https://pith.science/pith/7FHFSF23I6ESFULGQAF7UQXU2F/action/author_attestation","sign_citation":"https://pith.science/pith/7FHFSF23I6ESFULGQAF7UQXU2F/action/citation_signature","submit_replication":"https://pith.science/pith/7FHFSF23I6ESFULGQAF7UQXU2F/action/replication_record"}},"created_at":"2026-07-05T10:18:04.757133+00:00","updated_at":"2026-07-05T10:18:04.757133+00:00"}