{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:AOA5CRYTGYAWL3TOLVLRUFY54X","short_pith_number":"pith:AOA5CRYT","canonical_record":{"source":{"id":"2604.10027","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-04-11T04:49:11Z","cross_cats_sorted":[],"title_canon_sha256":"ba402a401de0d1f987d99155a85e176303d86d3f7eea795b622b9d17c2b8a8ee","abstract_canon_sha256":"9a6b7ffd67c7940fc0dbfaa9ae4eca0cba0eb505ff3e43d86d7c1c2f52926df2"},"schema_version":"1.0"},"canonical_sha256":"0381d14713360165ee6e5d571a171de5c55bb14cedd8d0a2dd427ec5971eaa16","source":{"kind":"arxiv","id":"2604.10027","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.10027","created_at":"2026-05-20T00:03:10Z"},{"alias_kind":"arxiv_version","alias_value":"2604.10027v2","created_at":"2026-05-20T00:03:10Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.10027","created_at":"2026-05-20T00:03:10Z"},{"alias_kind":"pith_short_12","alias_value":"AOA5CRYTGYAW","created_at":"2026-05-20T00:03:10Z"},{"alias_kind":"pith_short_16","alias_value":"AOA5CRYTGYAWL3TO","created_at":"2026-05-20T00:03:10Z"},{"alias_kind":"pith_short_8","alias_value":"AOA5CRYT","created_at":"2026-05-20T00:03:10Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:AOA5CRYTGYAWL3TOLVLRUFY54X","target":"record","payload":{"canonical_record":{"source":{"id":"2604.10027","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-04-11T04:49:11Z","cross_cats_sorted":[],"title_canon_sha256":"ba402a401de0d1f987d99155a85e176303d86d3f7eea795b622b9d17c2b8a8ee","abstract_canon_sha256":"9a6b7ffd67c7940fc0dbfaa9ae4eca0cba0eb505ff3e43d86d7c1c2f52926df2"},"schema_version":"1.0"},"canonical_sha256":"0381d14713360165ee6e5d571a171de5c55bb14cedd8d0a2dd427ec5971eaa16","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:03:10.952568Z","signature_b64":"HHAKSQvCdgWLXQszzvDCf0JuKOyp/2Thk+PpegX+GsgVb3dj0o9IvzVULZ8w4YVJyFif8dOzxCn1Ogjvk4xaDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0381d14713360165ee6e5d571a171de5c55bb14cedd8d0a2dd427ec5971eaa16","last_reissued_at":"2026-05-20T00:03:10.951599Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:03:10.951599Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2604.10027","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":"2026-05-20T00:03:10Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"zO3vnCIY+4Q+etMIT+9tlClrZu9T3gQVBZ+n5bBpbBvv2FjOdMwH4spwPIajPDcnc7gsHDPV2mEmgcNT0RIwCA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-09T02:37:26.286462Z"},"content_sha256":"676a7d4537c2c6507d022fa9ae1c7569860143751c47dc16a962cac91518708f","schema_version":"1.0","event_id":"sha256:676a7d4537c2c6507d022fa9ae1c7569860143751c47dc16a962cac91518708f"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:AOA5CRYTGYAWL3TOLVLRUFY54X","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"SinkTrack: Attention Sink based Context Anchoring for Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"By injecting contextual features into the attention sink token, SinkTrack keeps large language models focused on the original input throughout generation, reducing hallucination and forgetting.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Guikun Chen, Wenguan Wang, Xu Liu","submitted_at":"2026-04-11T04:49:11Z","abstract_excerpt":"Large language models (LLMs) suffer from hallucination and context forgetting. Prior studies suggest that attention drift is a primary cause of these problems, where LLMs' focus shifts towards newly generated tokens and away from the initial input context. To counteract this, we make use of a related, intrinsic characteristic of LLMs: attention sink -- the tendency to consistently allocate high attention to the very first token (i.e., <BOS>) of a sequence. Concretely, we propose an advanced context anchoring method, SinkTrack, which treats <BOS> as an information anchor and injects key context"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experiments demonstrate that SinkTrack mitigates hallucination and context forgetting across both textual (e.g., +21.6% on SQuAD2.0 with Llama3.1-8B-Instruct) and multi-modal (e.g., +22.8% on M3CoT with Qwen2.5-VL-7B-Instruct) tasks.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That consistently high attention to the BOS token can be reliably turned into an effective context anchor simply by injecting input-derived features into its representation, without side effects or the need for per-model tuning.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SinkTrack uses attention sink at the BOS token to anchor LLMs to initial context, reducing hallucination and forgetting with reported gains on benchmarks like SQuAD2.0 and M3CoT.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"By injecting contextual features into the attention sink token, SinkTrack keeps large language models focused on the original input throughout generation, reducing hallucination and forgetting.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"6e3013d3d0d18f75f2368e172817751f2fa04a37953db2bfa6d6cf945b83d036"},"source":{"id":"2604.10027","kind":"arxiv","version":2},"verdict":{"id":"6db75f4c-5a58-4212-be50-8d9f1fef30d3","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T15:43:49.317116Z","strongest_claim":"Experiments demonstrate that SinkTrack mitigates hallucination and context forgetting across both textual (e.g., +21.6% on SQuAD2.0 with Llama3.1-8B-Instruct) and multi-modal (e.g., +22.8% on M3CoT with Qwen2.5-VL-7B-Instruct) tasks.","one_line_summary":"SinkTrack uses attention sink at the BOS token to anchor LLMs to initial context, reducing hallucination and forgetting with reported gains on benchmarks like SQuAD2.0 and M3CoT.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That consistently high attention to the BOS token can be reliably turned into an effective context anchor simply by injecting input-derived features into its representation, without side effects or the need for per-model tuning.","pith_extraction_headline":"By injecting contextual features into the attention sink token, SinkTrack keeps large language models focused on the original input throughout generation, reducing hallucination and forgetting."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.10027/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":"6db75f4c-5a58-4212-be50-8d9f1fef30d3"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T00:03:10Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Y5gvziF3Eulf1Pyj9BDPK6F1sll4VZmamV2gxAJxWXDPpGyzFXZtQNkvv2W1GmJ00sGLvVzpwun76OjW2TwBBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-09T02:37:26.287379Z"},"content_sha256":"284c4133575774d3646f630bcba5fd6d069f322677a41c0604677b5995defb54","schema_version":"1.0","event_id":"sha256:284c4133575774d3646f630bcba5fd6d069f322677a41c0604677b5995defb54"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/AOA5CRYTGYAWL3TOLVLRUFY54X/bundle.json","state_url":"https://pith.science/pith/AOA5CRYTGYAWL3TOLVLRUFY54X/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/AOA5CRYTGYAWL3TOLVLRUFY54X/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":"2026-06-09T02:37:26Z","links":{"resolver":"https://pith.science/pith/AOA5CRYTGYAWL3TOLVLRUFY54X","bundle":"https://pith.science/pith/AOA5CRYTGYAWL3TOLVLRUFY54X/bundle.json","state":"https://pith.science/pith/AOA5CRYTGYAWL3TOLVLRUFY54X/state.json","well_known_bundle":"https://pith.science/.well-known/pith/AOA5CRYTGYAWL3TOLVLRUFY54X/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:AOA5CRYTGYAWL3TOLVLRUFY54X","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":"9a6b7ffd67c7940fc0dbfaa9ae4eca0cba0eb505ff3e43d86d7c1c2f52926df2","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-04-11T04:49:11Z","title_canon_sha256":"ba402a401de0d1f987d99155a85e176303d86d3f7eea795b622b9d17c2b8a8ee"},"schema_version":"1.0","source":{"id":"2604.10027","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.10027","created_at":"2026-05-20T00:03:10Z"},{"alias_kind":"arxiv_version","alias_value":"2604.10027v2","created_at":"2026-05-20T00:03:10Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.10027","created_at":"2026-05-20T00:03:10Z"},{"alias_kind":"pith_short_12","alias_value":"AOA5CRYTGYAW","created_at":"2026-05-20T00:03:10Z"},{"alias_kind":"pith_short_16","alias_value":"AOA5CRYTGYAWL3TO","created_at":"2026-05-20T00:03:10Z"},{"alias_kind":"pith_short_8","alias_value":"AOA5CRYT","created_at":"2026-05-20T00:03:10Z"}],"graph_snapshots":[{"event_id":"sha256:284c4133575774d3646f630bcba5fd6d069f322677a41c0604677b5995defb54","target":"graph","created_at":"2026-05-20T00:03:10Z","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":"Experiments demonstrate that SinkTrack mitigates hallucination and context forgetting across both textual (e.g., +21.6% on SQuAD2.0 with Llama3.1-8B-Instruct) and multi-modal (e.g., +22.8% on M3CoT with Qwen2.5-VL-7B-Instruct) tasks."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That consistently high attention to the BOS token can be reliably turned into an effective context anchor simply by injecting input-derived features into its representation, without side effects or the need for per-model tuning."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"SinkTrack uses attention sink at the BOS token to anchor LLMs to initial context, reducing hallucination and forgetting with reported gains on benchmarks like SQuAD2.0 and M3CoT."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"By injecting contextual features into the attention sink token, SinkTrack keeps large language models focused on the original input throughout generation, reducing hallucination and forgetting."}],"snapshot_sha256":"6e3013d3d0d18f75f2368e172817751f2fa04a37953db2bfa6d6cf945b83d036"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2604.10027/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Large language models (LLMs) suffer from hallucination and context forgetting. Prior studies suggest that attention drift is a primary cause of these problems, where LLMs' focus shifts towards newly generated tokens and away from the initial input context. To counteract this, we make use of a related, intrinsic characteristic of LLMs: attention sink -- the tendency to consistently allocate high attention to the very first token (i.e., <BOS>) of a sequence. Concretely, we propose an advanced context anchoring method, SinkTrack, which treats <BOS> as an information anchor and injects key context","authors_text":"Guikun Chen, Wenguan Wang, Xu Liu","cross_cats":[],"headline":"By injecting contextual features into the attention sink token, SinkTrack keeps large language models focused on the original input throughout generation, reducing hallucination and forgetting.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-04-11T04:49:11Z","title":"SinkTrack: Attention Sink based Context Anchoring for Large Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.10027","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-10T15:43:49.317116Z","id":"6db75f4c-5a58-4212-be50-8d9f1fef30d3","model_set":{"reader":"grok-4.3"},"one_line_summary":"SinkTrack uses attention sink at the BOS token to anchor LLMs to initial context, reducing hallucination and forgetting with reported gains on benchmarks like SQuAD2.0 and M3CoT.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"By injecting contextual features into the attention sink token, SinkTrack keeps large language models focused on the original input throughout generation, reducing hallucination and forgetting.","strongest_claim":"Experiments demonstrate that SinkTrack mitigates hallucination and context forgetting across both textual (e.g., +21.6% on SQuAD2.0 with Llama3.1-8B-Instruct) and multi-modal (e.g., +22.8% on M3CoT with Qwen2.5-VL-7B-Instruct) tasks.","weakest_assumption":"That consistently high attention to the BOS token can be reliably turned into an effective context anchor simply by injecting input-derived features into its representation, without side effects or the need for per-model tuning."}},"verdict_id":"6db75f4c-5a58-4212-be50-8d9f1fef30d3"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:676a7d4537c2c6507d022fa9ae1c7569860143751c47dc16a962cac91518708f","target":"record","created_at":"2026-05-20T00:03:10Z","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":"9a6b7ffd67c7940fc0dbfaa9ae4eca0cba0eb505ff3e43d86d7c1c2f52926df2","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-04-11T04:49:11Z","title_canon_sha256":"ba402a401de0d1f987d99155a85e176303d86d3f7eea795b622b9d17c2b8a8ee"},"schema_version":"1.0","source":{"id":"2604.10027","kind":"arxiv","version":2}},"canonical_sha256":"0381d14713360165ee6e5d571a171de5c55bb14cedd8d0a2dd427ec5971eaa16","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"0381d14713360165ee6e5d571a171de5c55bb14cedd8d0a2dd427ec5971eaa16","first_computed_at":"2026-05-20T00:03:10.951599Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:03:10.951599Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"HHAKSQvCdgWLXQszzvDCf0JuKOyp/2Thk+PpegX+GsgVb3dj0o9IvzVULZ8w4YVJyFif8dOzxCn1Ogjvk4xaDQ==","signature_status":"signed_v1","signed_at":"2026-05-20T00:03:10.952568Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.10027","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:676a7d4537c2c6507d022fa9ae1c7569860143751c47dc16a962cac91518708f","sha256:284c4133575774d3646f630bcba5fd6d069f322677a41c0604677b5995defb54"],"state_sha256":"ba07d144f54ed338e8279b58cd0cc47837526b28c13e4baafb5ab880eaaf8d26"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"a1hl7eeEZl1FjmlVflBvOEAm0IGwmhhX6iWL07wuIsNm4RcFq8jKpg0HDZktfMgM/nQT5BPCGhjfExE5ZF4+Dw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-09T02:37:26.291408Z","bundle_sha256":"dbc7fd57245340f1ce142fa986cc048fa88818e228cc275f69408b87035c648b"}}