{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:THY5BXRITUJAOUPAUJ74T6EL6D","short_pith_number":"pith:THY5BXRI","canonical_record":{"source":{"id":"2502.15007","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-20T19:59:35Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"def1824e5accd12d0ecf4e9b241e15337bce5255a9dcae4e0fbbe19e76259800","abstract_canon_sha256":"087daa7a2b841f63284d4b01664bf5a8e21ed978eb7258845310d480536178c5"},"schema_version":"1.0"},"canonical_sha256":"99f1d0de289d120751e0a27fc9f88bf0c9f23737d37d20459af9e56d86eb5e79","source":{"kind":"arxiv","id":"2502.15007","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2502.15007","created_at":"2026-07-05T10:17:37Z"},{"alias_kind":"arxiv_version","alias_value":"2502.15007v1","created_at":"2026-07-05T10:17:37Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.15007","created_at":"2026-07-05T10:17:37Z"},{"alias_kind":"pith_short_12","alias_value":"THY5BXRITUJA","created_at":"2026-07-05T10:17:37Z"},{"alias_kind":"pith_short_16","alias_value":"THY5BXRITUJAOUPA","created_at":"2026-07-05T10:17:37Z"},{"alias_kind":"pith_short_8","alias_value":"THY5BXRI","created_at":"2026-07-05T10:17:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:THY5BXRITUJAOUPAUJ74T6EL6D","target":"record","payload":{"canonical_record":{"source":{"id":"2502.15007","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-20T19:59:35Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"def1824e5accd12d0ecf4e9b241e15337bce5255a9dcae4e0fbbe19e76259800","abstract_canon_sha256":"087daa7a2b841f63284d4b01664bf5a8e21ed978eb7258845310d480536178c5"},"schema_version":"1.0"},"canonical_sha256":"99f1d0de289d120751e0a27fc9f88bf0c9f23737d37d20459af9e56d86eb5e79","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:17:37.999813Z","signature_b64":"H3WUu/Nn1veB5gG0U+V6EjBolaavW62OlCE6cjW1oeWUWTUHhPewLAPv56AB2V6CRlmCqfaFJi/3N9D2E+LFDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"99f1d0de289d120751e0a27fc9f88bf0c9f23737d37d20459af9e56d86eb5e79","last_reissued_at":"2026-07-05T10:17:37.999329Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:17:37.999329Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2502.15007","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":"2026-07-05T10:17:37Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"VM0m0FtHU4zQUd9jABsI7sFA2qJH3bFBm/3IItznnfcvbbtYH3snEP5RhODriF0nzIMZ9xsBkGCXiBPq7hePAg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-31T17:46:28.763733Z"},"content_sha256":"41a15b6115d8564062998d8a51651848be82a7e3de1a6ad3fc21275305216ffd","schema_version":"1.0","event_id":"sha256:41a15b6115d8564062998d8a51651848be82a7e3de1a6ad3fc21275305216ffd"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:THY5BXRITUJAOUPAUJ74T6EL6D","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"LLM-Microscope: Uncovering the Hidden Role of Punctuation in Context Memory of Transformers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Andrey Kuznetsov, Anton Razzhigaev, Elizaveta Goncharova, Ivan Oseledets, Matvey Mikhalchuk, Polina Druzhinina, Temurbek Rahmatullaev","submitted_at":"2025-02-20T19:59:35Z","abstract_excerpt":"We introduce methods to quantify how Large Language Models (LLMs) encode and store contextual information, revealing that tokens often seen as minor (e.g., determiners, punctuation) carry surprisingly high context. Notably, removing these tokens -- especially stopwords, articles, and commas -- consistently degrades performance on MMLU and BABILong-4k, even if removing only irrelevant tokens. Our analysis also shows a strong correlation between contextualization and linearity, where linearity measures how closely the transformation from one layer's embeddings to the next can be approximated by "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.15007","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/2502.15007/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":"2026-07-05T10:17:37Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"QR6Xh+xRNHdGssCHMr/cP1b8mDhu9EB2LQjkERBr6Vd2gMfTlhhMTtAK2Zzgc8BcI52vJnGVTgabJNOjNcieBw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-31T17:46:28.764183Z"},"content_sha256":"7c10628d2a6b259f39756c545e09237c221b71950e96ef43f45f8a430522cf85","schema_version":"1.0","event_id":"sha256:7c10628d2a6b259f39756c545e09237c221b71950e96ef43f45f8a430522cf85"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/THY5BXRITUJAOUPAUJ74T6EL6D/bundle.json","state_url":"https://pith.science/pith/THY5BXRITUJAOUPAUJ74T6EL6D/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/THY5BXRITUJAOUPAUJ74T6EL6D/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":"2026-07-31T17:46:28Z","links":{"resolver":"https://pith.science/pith/THY5BXRITUJAOUPAUJ74T6EL6D","bundle":"https://pith.science/pith/THY5BXRITUJAOUPAUJ74T6EL6D/bundle.json","state":"https://pith.science/pith/THY5BXRITUJAOUPAUJ74T6EL6D/state.json","well_known_bundle":"https://pith.science/.well-known/pith/THY5BXRITUJAOUPAUJ74T6EL6D/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:THY5BXRITUJAOUPAUJ74T6EL6D","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":"087daa7a2b841f63284d4b01664bf5a8e21ed978eb7258845310d480536178c5","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-20T19:59:35Z","title_canon_sha256":"def1824e5accd12d0ecf4e9b241e15337bce5255a9dcae4e0fbbe19e76259800"},"schema_version":"1.0","source":{"id":"2502.15007","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2502.15007","created_at":"2026-07-05T10:17:37Z"},{"alias_kind":"arxiv_version","alias_value":"2502.15007v1","created_at":"2026-07-05T10:17:37Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.15007","created_at":"2026-07-05T10:17:37Z"},{"alias_kind":"pith_short_12","alias_value":"THY5BXRITUJA","created_at":"2026-07-05T10:17:37Z"},{"alias_kind":"pith_short_16","alias_value":"THY5BXRITUJAOUPA","created_at":"2026-07-05T10:17:37Z"},{"alias_kind":"pith_short_8","alias_value":"THY5BXRI","created_at":"2026-07-05T10:17:37Z"}],"graph_snapshots":[{"event_id":"sha256:7c10628d2a6b259f39756c545e09237c221b71950e96ef43f45f8a430522cf85","target":"graph","created_at":"2026-07-05T10:17:37Z","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/2502.15007/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"We introduce methods to quantify how Large Language Models (LLMs) encode and store contextual information, revealing that tokens often seen as minor (e.g., determiners, punctuation) carry surprisingly high context. Notably, removing these tokens -- especially stopwords, articles, and commas -- consistently degrades performance on MMLU and BABILong-4k, even if removing only irrelevant tokens. Our analysis also shows a strong correlation between contextualization and linearity, where linearity measures how closely the transformation from one layer's embeddings to the next can be approximated by ","authors_text":"Andrey Kuznetsov, Anton Razzhigaev, Elizaveta Goncharova, Ivan Oseledets, Matvey Mikhalchuk, Polina Druzhinina, Temurbek Rahmatullaev","cross_cats":["cs.AI"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-20T19:59:35Z","title":"LLM-Microscope: Uncovering the Hidden Role of Punctuation in Context Memory of Transformers"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.15007","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":"sha256:41a15b6115d8564062998d8a51651848be82a7e3de1a6ad3fc21275305216ffd","target":"record","created_at":"2026-07-05T10:17:37Z","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":"087daa7a2b841f63284d4b01664bf5a8e21ed978eb7258845310d480536178c5","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-20T19:59:35Z","title_canon_sha256":"def1824e5accd12d0ecf4e9b241e15337bce5255a9dcae4e0fbbe19e76259800"},"schema_version":"1.0","source":{"id":"2502.15007","kind":"arxiv","version":1}},"canonical_sha256":"99f1d0de289d120751e0a27fc9f88bf0c9f23737d37d20459af9e56d86eb5e79","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"99f1d0de289d120751e0a27fc9f88bf0c9f23737d37d20459af9e56d86eb5e79","first_computed_at":"2026-07-05T10:17:37.999329Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:17:37.999329Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"H3WUu/Nn1veB5gG0U+V6EjBolaavW62OlCE6cjW1oeWUWTUHhPewLAPv56AB2V6CRlmCqfaFJi/3N9D2E+LFDQ==","signature_status":"signed_v1","signed_at":"2026-07-05T10:17:37.999813Z","signed_message":"canonical_sha256_bytes"},"source_id":"2502.15007","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:41a15b6115d8564062998d8a51651848be82a7e3de1a6ad3fc21275305216ffd","sha256:7c10628d2a6b259f39756c545e09237c221b71950e96ef43f45f8a430522cf85"],"state_sha256":"7812cdfc61efc71a3201019a49387784eb6362e35f17e2409b9658be46abbb63"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"WlZr9+6aymB7lYkefA9LFnlBIRdhc4iqHfUx2iYojL2jn+Cwe6xpFWMB41AkZVwvGiXHb7ikZaYGU8Cu6YR7Cg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-31T17:46:28.768532Z","bundle_sha256":"941a0662fabb66d2e8f341f6136706fcc95875823d2025860e63d8ef0ec588e1"}}