{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:GGSRFUKH77XNVTRKKA6WY7RVVP","short_pith_number":"pith:GGSRFUKH","canonical_record":{"source":{"id":"2607.25818","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-07-28T15:01:02Z","cross_cats_sorted":[],"title_canon_sha256":"48b45b4e2c2d4b2f0ce8f5f3d15cf2f353d9ea365f385614b14c4ebefeda5207","abstract_canon_sha256":"220e4ec9f4a44e7fc21e666dacf2e4644199ee50d57559effa556cb176098f3b"},"schema_version":"1.0"},"canonical_sha256":"31a512d147ffeedace2a503d6c7e35abd2de5c35c4abbe7bad0b6c1d82ac5da6","source":{"kind":"arxiv","id":"2607.25818","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2607.25818","created_at":"2026-07-29T01:26:04Z"},{"alias_kind":"arxiv_version","alias_value":"2607.25818v1","created_at":"2026-07-29T01:26:04Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.25818","created_at":"2026-07-29T01:26:04Z"},{"alias_kind":"pith_short_12","alias_value":"GGSRFUKH77XN","created_at":"2026-07-29T01:26:04Z"},{"alias_kind":"pith_short_16","alias_value":"GGSRFUKH77XNVTRK","created_at":"2026-07-29T01:26:04Z"},{"alias_kind":"pith_short_8","alias_value":"GGSRFUKH","created_at":"2026-07-29T01:26:04Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:GGSRFUKH77XNVTRKKA6WY7RVVP","target":"record","payload":{"canonical_record":{"source":{"id":"2607.25818","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-07-28T15:01:02Z","cross_cats_sorted":[],"title_canon_sha256":"48b45b4e2c2d4b2f0ce8f5f3d15cf2f353d9ea365f385614b14c4ebefeda5207","abstract_canon_sha256":"220e4ec9f4a44e7fc21e666dacf2e4644199ee50d57559effa556cb176098f3b"},"schema_version":"1.0"},"canonical_sha256":"31a512d147ffeedace2a503d6c7e35abd2de5c35c4abbe7bad0b6c1d82ac5da6","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-29T01:26:04.514781Z","signature_b64":"MyG0l2CZm19A1Ox0BcHiB9q/P5o3wKXU1+iRKAHhYkO6GqD6dAPXE/2rYTvwU43tD4+hzgCzRU1PEBskSTCpDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"31a512d147ffeedace2a503d6c7e35abd2de5c35c4abbe7bad0b6c1d82ac5da6","last_reissued_at":"2026-07-29T01:26:04.513846Z","signature_status":"signed_v1","first_computed_at":"2026-07-29T01:26:04.513846Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2607.25818","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":"2026-07-29T01:26:04Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"6341sKHMOGIn9tLUOVbtdF6qHhY4ayrwNaTt8hB6U7AzxwOFRtLynDu7xbjhbbcOSw5aIXsskI8n5C57FaFPAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-18T12:40:53.447068Z"},"content_sha256":"2a777619a3e9969aea8ec4784f71bb1a022e1973c15cad46866203cafe375444","schema_version":"1.0","event_id":"sha256:2a777619a3e9969aea8ec4784f71bb1a022e1973c15cad46866203cafe375444"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:GGSRFUKH77XNVTRKKA6WY7RVVP","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"SepPrune:A Separator-based Pruning Framework for Efficient Multimodal Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Qihui Zhu, Siying Wu, Xiaoyan Sun, Yang Liu, Yuchen Wang","submitted_at":"2026-07-28T15:01:02Z","abstract_excerpt":"Recent multimodal large language models (MLLMs), such as Qwen2.5-VL and InternVL3, generate large numbers of vision tokens for high-resolution inputs, leading to substantial computational cost. Existing vision token pruning methods either depend on cross-modal attention and cannot prune before the prefill stage, or rely on diversity estimation with high computational overhead. We observe that attention scores from both vision and text tokens peak at modality separator tokens, suggesting that these separators bridge the two modalities. Based on this observation, we propose SepPrune, an efficien"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.25818","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/2607.25818/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":"2026-07-29T01:26:04Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"zuMQ9AI4MQUdHiHnoUFqvakv/6o2qi9apRvebmXtxtOhN1miGryjLBgUwTTToP8VP4XEWtyjj4wrC0EwosNXBw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-18T12:40:53.447940Z"},"content_sha256":"e126905605bbdce652c0c172ea2a21e3f06908a437f5fea2f68dfc93e990744f","schema_version":"1.0","event_id":"sha256:e126905605bbdce652c0c172ea2a21e3f06908a437f5fea2f68dfc93e990744f"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/GGSRFUKH77XNVTRKKA6WY7RVVP/bundle.json","state_url":"https://pith.science/pith/GGSRFUKH77XNVTRKKA6WY7RVVP/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/GGSRFUKH77XNVTRKKA6WY7RVVP/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":"2026-08-18T12:40:53Z","links":{"resolver":"https://pith.science/pith/GGSRFUKH77XNVTRKKA6WY7RVVP","bundle":"https://pith.science/pith/GGSRFUKH77XNVTRKKA6WY7RVVP/bundle.json","state":"https://pith.science/pith/GGSRFUKH77XNVTRKKA6WY7RVVP/state.json","well_known_bundle":"https://pith.science/.well-known/pith/GGSRFUKH77XNVTRKKA6WY7RVVP/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:GGSRFUKH77XNVTRKKA6WY7RVVP","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":"220e4ec9f4a44e7fc21e666dacf2e4644199ee50d57559effa556cb176098f3b","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-07-28T15:01:02Z","title_canon_sha256":"48b45b4e2c2d4b2f0ce8f5f3d15cf2f353d9ea365f385614b14c4ebefeda5207"},"schema_version":"1.0","source":{"id":"2607.25818","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2607.25818","created_at":"2026-07-29T01:26:04Z"},{"alias_kind":"arxiv_version","alias_value":"2607.25818v1","created_at":"2026-07-29T01:26:04Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.25818","created_at":"2026-07-29T01:26:04Z"},{"alias_kind":"pith_short_12","alias_value":"GGSRFUKH77XN","created_at":"2026-07-29T01:26:04Z"},{"alias_kind":"pith_short_16","alias_value":"GGSRFUKH77XNVTRK","created_at":"2026-07-29T01:26:04Z"},{"alias_kind":"pith_short_8","alias_value":"GGSRFUKH","created_at":"2026-07-29T01:26:04Z"}],"graph_snapshots":[{"event_id":"sha256:e126905605bbdce652c0c172ea2a21e3f06908a437f5fea2f68dfc93e990744f","target":"graph","created_at":"2026-07-29T01:26:04Z","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/2607.25818/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Recent multimodal large language models (MLLMs), such as Qwen2.5-VL and InternVL3, generate large numbers of vision tokens for high-resolution inputs, leading to substantial computational cost. Existing vision token pruning methods either depend on cross-modal attention and cannot prune before the prefill stage, or rely on diversity estimation with high computational overhead. We observe that attention scores from both vision and text tokens peak at modality separator tokens, suggesting that these separators bridge the two modalities. Based on this observation, we propose SepPrune, an efficien","authors_text":"Qihui Zhu, Siying Wu, Xiaoyan Sun, Yang Liu, Yuchen Wang","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-07-28T15:01:02Z","title":"SepPrune:A Separator-based Pruning Framework for Efficient Multimodal Large Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.25818","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":"sha256:2a777619a3e9969aea8ec4784f71bb1a022e1973c15cad46866203cafe375444","target":"record","created_at":"2026-07-29T01:26:04Z","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":"220e4ec9f4a44e7fc21e666dacf2e4644199ee50d57559effa556cb176098f3b","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-07-28T15:01:02Z","title_canon_sha256":"48b45b4e2c2d4b2f0ce8f5f3d15cf2f353d9ea365f385614b14c4ebefeda5207"},"schema_version":"1.0","source":{"id":"2607.25818","kind":"arxiv","version":1}},"canonical_sha256":"31a512d147ffeedace2a503d6c7e35abd2de5c35c4abbe7bad0b6c1d82ac5da6","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"31a512d147ffeedace2a503d6c7e35abd2de5c35c4abbe7bad0b6c1d82ac5da6","first_computed_at":"2026-07-29T01:26:04.513846Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-29T01:26:04.513846Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"MyG0l2CZm19A1Ox0BcHiB9q/P5o3wKXU1+iRKAHhYkO6GqD6dAPXE/2rYTvwU43tD4+hzgCzRU1PEBskSTCpDQ==","signature_status":"signed_v1","signed_at":"2026-07-29T01:26:04.514781Z","signed_message":"canonical_sha256_bytes"},"source_id":"2607.25818","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:2a777619a3e9969aea8ec4784f71bb1a022e1973c15cad46866203cafe375444","sha256:e126905605bbdce652c0c172ea2a21e3f06908a437f5fea2f68dfc93e990744f"],"state_sha256":"0f800226c1f27e00192d2906d63589ecd4fae72eb1c6f4f3a0b1b0f11a12ca47"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"OL3+mqM9D4J99k5F1HyizMrNKO434/kGvKdG1t43bHrC0wevtTN4gvIKqazXBLRpuZ4WJBk2Ltbc+77Z85/WBw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-18T12:40:53.453622Z","bundle_sha256":"59f74b64a23b2294709d828f162339e1468ac8274b8d3130a2ee5da0a4c426e2"}}