{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:KTUGVVS2PXHK6RH54FW5B2KZ2U","short_pith_number":"pith:KTUGVVS2","canonical_record":{"source":{"id":"2604.18103","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-04-20T11:20:03Z","cross_cats_sorted":[],"title_canon_sha256":"c5396db11b831878e98a5f45538547177fe7196950cc64aa19519f3b1be9e01d","abstract_canon_sha256":"b25ebb6577b94dc4282a0a63d264112298752b99cf29ae14a6971bfd1a2b5e1e"},"schema_version":"1.0"},"canonical_sha256":"54e86ad65a7dceaf44fde16dd0e959d529d77458ecc718da3596c88a58cd4238","source":{"kind":"arxiv","id":"2604.18103","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.18103","created_at":"2026-05-27T01:05:54Z"},{"alias_kind":"arxiv_version","alias_value":"2604.18103v2","created_at":"2026-05-27T01:05:54Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.18103","created_at":"2026-05-27T01:05:54Z"},{"alias_kind":"pith_short_12","alias_value":"KTUGVVS2PXHK","created_at":"2026-05-27T01:05:54Z"},{"alias_kind":"pith_short_16","alias_value":"KTUGVVS2PXHK6RH5","created_at":"2026-05-27T01:05:54Z"},{"alias_kind":"pith_short_8","alias_value":"KTUGVVS2","created_at":"2026-05-27T01:05:54Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:KTUGVVS2PXHK6RH54FW5B2KZ2U","target":"record","payload":{"canonical_record":{"source":{"id":"2604.18103","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-04-20T11:20:03Z","cross_cats_sorted":[],"title_canon_sha256":"c5396db11b831878e98a5f45538547177fe7196950cc64aa19519f3b1be9e01d","abstract_canon_sha256":"b25ebb6577b94dc4282a0a63d264112298752b99cf29ae14a6971bfd1a2b5e1e"},"schema_version":"1.0"},"canonical_sha256":"54e86ad65a7dceaf44fde16dd0e959d529d77458ecc718da3596c88a58cd4238","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-27T01:05:54.822487Z","signature_b64":"Zgj9jsl6muyxPXD5vYRFuCHtdrdL6UEMX5EBM4zL/n9Z663QWX0Fw1MlbGMokYO44SvlRWkth55CCAKyyeuABw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"54e86ad65a7dceaf44fde16dd0e959d529d77458ecc718da3596c88a58cd4238","last_reissued_at":"2026-05-27T01:05:54.821597Z","signature_status":"signed_v1","first_computed_at":"2026-05-27T01:05:54.821597Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2604.18103","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-27T01:05:54Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"q4+ugVB0WwrZ5gRQofFdyG+gVBMGFSksj4nM+fYRxsBkuqT9I2iwnS1eBLKd9Wm+wGxfGBMxyYPLaHP5tLrlAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-27T18:57:09.496029Z"},"content_sha256":"1dc689435ec1274cd41021312b4640067754baecfdd5d924c8ca323a47f743c1","schema_version":"1.0","event_id":"sha256:1dc689435ec1274cd41021312b4640067754baecfdd5d924c8ca323a47f743c1"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:KTUGVVS2PXHK6RH54FW5B2KZ2U","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Stability Implies Redundancy: Delta Attention Selective Halting for Efficient Long-Context Prefilling","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Tokens reach semantic fixing points where further attention adds nothing, so early halting cuts prefill cost.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Linfeng Zhang, Shaobo Wang, Tailai Chen, Yifeng Gao, Yijue Xu, Yujie Chen, Zoe Wanying He","submitted_at":"2026-04-20T11:20:03Z","abstract_excerpt":"Prefilling computational costs pose a significant bottleneck for Large Language Models (LLMs) and Large Multimodal Models (LMMs) in long-context settings. While token pruning reduces sequence length, prior methods rely on heuristics that break compatibility with hardware-efficient kernels like FlashAttention. In this work, we observe that tokens evolve toward \\textit{semantic fixing points}, making further processing redundant. To this end, we introduce Delta Attention Selective Halting (DASH), a training-free policy that monitors the layer-wise update dynamics of the self-attention mechanism "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Tokens evolve toward semantic fixing points, making further processing redundant; DASH monitors layer-wise update dynamics of self-attention to selectively halt stabilized tokens and delivers significant prefill speedups while preserving accuracy and hardware efficiency.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That monitoring layer-wise self-attention deltas reliably identifies tokens whose further computation is redundant without discarding task-critical information, and that this holds across models and tasks without per-model tuning.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"DASH selectively halts stabilized tokens by monitoring layer-wise self-attention deltas, delivering prefill speedups on language and vision tasks without accuracy loss or hardware incompatibility.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Tokens reach semantic fixing points where further attention adds nothing, so early halting cuts prefill cost.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"9d25ec1061c50d9d1d8d50cb162bf4d9174cf6f11f8eafdd068f7cb3d51cbc90"},"source":{"id":"2604.18103","kind":"arxiv","version":2},"verdict":{"id":"716d62aa-0f78-449b-b9c6-f09643d52732","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T04:37:00.311740Z","strongest_claim":"Tokens evolve toward semantic fixing points, making further processing redundant; DASH monitors layer-wise update dynamics of self-attention to selectively halt stabilized tokens and delivers significant prefill speedups while preserving accuracy and hardware efficiency.","one_line_summary":"DASH selectively halts stabilized tokens by monitoring layer-wise self-attention deltas, delivering prefill speedups on language and vision tasks without accuracy loss or hardware incompatibility.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That monitoring layer-wise self-attention deltas reliably identifies tokens whose further computation is redundant without discarding task-critical information, and that this holds across models and tasks without per-model tuning.","pith_extraction_headline":"Tokens reach semantic fixing points where further attention adds nothing, so early halting cuts prefill cost."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.18103/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_compliance","ran_at":"2026-05-20T04:26:08.555427Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"33e06639ba6363a14fc96f891c03f2d45d6032ca657c88c6b4dab87e38ff7a2f"},"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":"716d62aa-0f78-449b-b9c6-f09643d52732"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-27T01:05:54Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ssFlE7jUUbEcuOz3Xz/K7zodcGmRV646bW7w11Pv8d1ypPotu3DlN2cgr0seju03yzo7DebWJSNOw38u6mHiAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-27T18:57:09.496768Z"},"content_sha256":"4d169f78757e2aa5a6612fe513365f1891fd27cf8dec429af4c2fb9dffea45f6","schema_version":"1.0","event_id":"sha256:4d169f78757e2aa5a6612fe513365f1891fd27cf8dec429af4c2fb9dffea45f6"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/KTUGVVS2PXHK6RH54FW5B2KZ2U/bundle.json","state_url":"https://pith.science/pith/KTUGVVS2PXHK6RH54FW5B2KZ2U/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/KTUGVVS2PXHK6RH54FW5B2KZ2U/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":"2026-05-27T18:57:09Z","links":{"resolver":"https://pith.science/pith/KTUGVVS2PXHK6RH54FW5B2KZ2U","bundle":"https://pith.science/pith/KTUGVVS2PXHK6RH54FW5B2KZ2U/bundle.json","state":"https://pith.science/pith/KTUGVVS2PXHK6RH54FW5B2KZ2U/state.json","well_known_bundle":"https://pith.science/.well-known/pith/KTUGVVS2PXHK6RH54FW5B2KZ2U/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:KTUGVVS2PXHK6RH54FW5B2KZ2U","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":"b25ebb6577b94dc4282a0a63d264112298752b99cf29ae14a6971bfd1a2b5e1e","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-04-20T11:20:03Z","title_canon_sha256":"c5396db11b831878e98a5f45538547177fe7196950cc64aa19519f3b1be9e01d"},"schema_version":"1.0","source":{"id":"2604.18103","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.18103","created_at":"2026-05-27T01:05:54Z"},{"alias_kind":"arxiv_version","alias_value":"2604.18103v2","created_at":"2026-05-27T01:05:54Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.18103","created_at":"2026-05-27T01:05:54Z"},{"alias_kind":"pith_short_12","alias_value":"KTUGVVS2PXHK","created_at":"2026-05-27T01:05:54Z"},{"alias_kind":"pith_short_16","alias_value":"KTUGVVS2PXHK6RH5","created_at":"2026-05-27T01:05:54Z"},{"alias_kind":"pith_short_8","alias_value":"KTUGVVS2","created_at":"2026-05-27T01:05:54Z"}],"graph_snapshots":[{"event_id":"sha256:4d169f78757e2aa5a6612fe513365f1891fd27cf8dec429af4c2fb9dffea45f6","target":"graph","created_at":"2026-05-27T01:05:54Z","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":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Tokens evolve toward semantic fixing points, making further processing redundant; DASH monitors layer-wise update dynamics of self-attention to selectively halt stabilized tokens and delivers significant prefill speedups while preserving accuracy and hardware efficiency."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That monitoring layer-wise self-attention deltas reliably identifies tokens whose further computation is redundant without discarding task-critical information, and that this holds across models and tasks without per-model tuning."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"DASH selectively halts stabilized tokens by monitoring layer-wise self-attention deltas, delivering prefill speedups on language and vision tasks without accuracy loss or hardware incompatibility."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Tokens reach semantic fixing points where further attention adds nothing, so early halting cuts prefill cost."}],"snapshot_sha256":"9d25ec1061c50d9d1d8d50cb162bf4d9174cf6f11f8eafdd068f7cb3d51cbc90"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-20T04:26:08.555427Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2604.18103/integrity.json","findings":[],"snapshot_sha256":"33e06639ba6363a14fc96f891c03f2d45d6032ca657c88c6b4dab87e38ff7a2f","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Prefilling computational costs pose a significant bottleneck for Large Language Models (LLMs) and Large Multimodal Models (LMMs) in long-context settings. While token pruning reduces sequence length, prior methods rely on heuristics that break compatibility with hardware-efficient kernels like FlashAttention. In this work, we observe that tokens evolve toward \\textit{semantic fixing points}, making further processing redundant. To this end, we introduce Delta Attention Selective Halting (DASH), a training-free policy that monitors the layer-wise update dynamics of the self-attention mechanism ","authors_text":"Linfeng Zhang, Shaobo Wang, Tailai Chen, Yifeng Gao, Yijue Xu, Yujie Chen, Zoe Wanying He","cross_cats":[],"headline":"Tokens reach semantic fixing points where further attention adds nothing, so early halting cuts prefill cost.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-04-20T11:20:03Z","title":"Stability Implies Redundancy: Delta Attention Selective Halting for Efficient Long-Context Prefilling"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.18103","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-10T04:37:00.311740Z","id":"716d62aa-0f78-449b-b9c6-f09643d52732","model_set":{"reader":"grok-4.3"},"one_line_summary":"DASH selectively halts stabilized tokens by monitoring layer-wise self-attention deltas, delivering prefill speedups on language and vision tasks without accuracy loss or hardware incompatibility.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Tokens reach semantic fixing points where further attention adds nothing, so early halting cuts prefill cost.","strongest_claim":"Tokens evolve toward semantic fixing points, making further processing redundant; DASH monitors layer-wise update dynamics of self-attention to selectively halt stabilized tokens and delivers significant prefill speedups while preserving accuracy and hardware efficiency.","weakest_assumption":"That monitoring layer-wise self-attention deltas reliably identifies tokens whose further computation is redundant without discarding task-critical information, and that this holds across models and tasks without per-model tuning."}},"verdict_id":"716d62aa-0f78-449b-b9c6-f09643d52732"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:1dc689435ec1274cd41021312b4640067754baecfdd5d924c8ca323a47f743c1","target":"record","created_at":"2026-05-27T01:05:54Z","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":"b25ebb6577b94dc4282a0a63d264112298752b99cf29ae14a6971bfd1a2b5e1e","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-04-20T11:20:03Z","title_canon_sha256":"c5396db11b831878e98a5f45538547177fe7196950cc64aa19519f3b1be9e01d"},"schema_version":"1.0","source":{"id":"2604.18103","kind":"arxiv","version":2}},"canonical_sha256":"54e86ad65a7dceaf44fde16dd0e959d529d77458ecc718da3596c88a58cd4238","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"54e86ad65a7dceaf44fde16dd0e959d529d77458ecc718da3596c88a58cd4238","first_computed_at":"2026-05-27T01:05:54.821597Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-27T01:05:54.821597Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"Zgj9jsl6muyxPXD5vYRFuCHtdrdL6UEMX5EBM4zL/n9Z663QWX0Fw1MlbGMokYO44SvlRWkth55CCAKyyeuABw==","signature_status":"signed_v1","signed_at":"2026-05-27T01:05:54.822487Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.18103","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:1dc689435ec1274cd41021312b4640067754baecfdd5d924c8ca323a47f743c1","sha256:4d169f78757e2aa5a6612fe513365f1891fd27cf8dec429af4c2fb9dffea45f6"],"state_sha256":"a1214a6b36cd21a047075fab01a45330c54ade6051f6745b9f77a109f726d9a3"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"vLS0Fs2Cv29U45Sc7/TvRKVVUYVKTMfz153dVbF+u/jHYvyzxcxgB/xa8EM3+y8zWHFZbUyW8v8aFEY/wpC3Aw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-27T18:57:09.499913Z","bundle_sha256":"f6aedc3d7edf26c86b954d25de7e63ac37babefdeae9547650bcbf338597d186"}}