{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:3J7SZEDFLDVDWUSO6VYGCG2LYI","short_pith_number":"pith:3J7SZEDF","canonical_record":{"source":{"id":"2501.06638","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-11T21:03:22Z","cross_cats_sorted":[],"title_canon_sha256":"8f2d8ba2c309f8bbf227b18a6caa403552fbbeb164668f8e08daaba0c8cd4e91","abstract_canon_sha256":"ea76cd5387087a1763abfb99bb0b654e834db41f65eb7ec9a05750b9e1bb9c41"},"schema_version":"1.0"},"canonical_sha256":"da7f2c906558ea3b524ef570611b4bc20691eddc24b03b0a828d3f174edc3be5","source":{"kind":"arxiv","id":"2501.06638","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2501.06638","created_at":"2026-07-05T10:00:07Z"},{"alias_kind":"arxiv_version","alias_value":"2501.06638v1","created_at":"2026-07-05T10:00:07Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.06638","created_at":"2026-07-05T10:00:07Z"},{"alias_kind":"pith_short_12","alias_value":"3J7SZEDFLDVD","created_at":"2026-07-05T10:00:07Z"},{"alias_kind":"pith_short_16","alias_value":"3J7SZEDFLDVDWUSO","created_at":"2026-07-05T10:00:07Z"},{"alias_kind":"pith_short_8","alias_value":"3J7SZEDF","created_at":"2026-07-05T10:00:07Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:3J7SZEDFLDVDWUSO6VYGCG2LYI","target":"record","payload":{"canonical_record":{"source":{"id":"2501.06638","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-11T21:03:22Z","cross_cats_sorted":[],"title_canon_sha256":"8f2d8ba2c309f8bbf227b18a6caa403552fbbeb164668f8e08daaba0c8cd4e91","abstract_canon_sha256":"ea76cd5387087a1763abfb99bb0b654e834db41f65eb7ec9a05750b9e1bb9c41"},"schema_version":"1.0"},"canonical_sha256":"da7f2c906558ea3b524ef570611b4bc20691eddc24b03b0a828d3f174edc3be5","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:00:07.631403Z","signature_b64":"DN5ajRQyOB3iFnMLx8q5wqfqpZqkj1x1+HNJMoO/A+BiH9G6cxfZgohQ8GlfN03qU4bRIpzVT/atDF30hKYMCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"da7f2c906558ea3b524ef570611b4bc20691eddc24b03b0a828d3f174edc3be5","last_reissued_at":"2026-07-05T10:00:07.630966Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:00:07.630966Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2501.06638","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":"2026-07-05T10:00:07Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"khP/lrRmBVWyla9XRKNY9ZPBVt3Okfs76d/fSdLjckh2KY92iLXuxpvwGBpK5UxwOqlrgK5G6BIIjbYptx+JAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-11T19:41:50.008708Z"},"content_sha256":"ef4a605591a8be8b4fa5f215189aadd631b64cf09b14d4b96b2ea44b617aa4fe","schema_version":"1.0","event_id":"sha256:ef4a605591a8be8b4fa5f215189aadd631b64cf09b14d4b96b2ea44b617aa4fe"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:3J7SZEDFLDVDWUSO6VYGCG2LYI","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Scaling Down Semantic Leakage: Investigating Associative Bias in Smaller Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Veronika Smilga","submitted_at":"2025-01-11T21:03:22Z","abstract_excerpt":"Semantic leakage is a phenomenon recently introduced by Gonen et al. (2024). It refers to a situation in which associations learnt from the training data emerge in language model generations in an unexpected and sometimes undesired way. Prior work has focused on leakage in large language models (7B+ parameters). In this study, I use Qwen2.5 model family to explore whether smaller models, ranging from 500M to 7B parameters, demonstrate less semantic leakage due to their limited capacity for capturing complex associations. Building on the previous dataset from Gonen et al. (2024), I introduce a "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.06638","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/2501.06638/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":"2026-07-05T10:00:07Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"6beG41GHtnyikqn1MjsXqRIxQjSg6XdgTf5PaTis5CYMZEx8LnB5tHMwvCktQoNdSDuRN+8rYDhb4yYW62EaAg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-11T19:41:50.009752Z"},"content_sha256":"bafb3d2bdb0345d8331e7c6781f6332587b2c6fffc57cf72f70fa23c4a8dbee2","schema_version":"1.0","event_id":"sha256:bafb3d2bdb0345d8331e7c6781f6332587b2c6fffc57cf72f70fa23c4a8dbee2"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/3J7SZEDFLDVDWUSO6VYGCG2LYI/bundle.json","state_url":"https://pith.science/pith/3J7SZEDFLDVDWUSO6VYGCG2LYI/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/3J7SZEDFLDVDWUSO6VYGCG2LYI/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":"2026-08-11T19:41:50Z","links":{"resolver":"https://pith.science/pith/3J7SZEDFLDVDWUSO6VYGCG2LYI","bundle":"https://pith.science/pith/3J7SZEDFLDVDWUSO6VYGCG2LYI/bundle.json","state":"https://pith.science/pith/3J7SZEDFLDVDWUSO6VYGCG2LYI/state.json","well_known_bundle":"https://pith.science/.well-known/pith/3J7SZEDFLDVDWUSO6VYGCG2LYI/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:3J7SZEDFLDVDWUSO6VYGCG2LYI","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":"ea76cd5387087a1763abfb99bb0b654e834db41f65eb7ec9a05750b9e1bb9c41","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-11T21:03:22Z","title_canon_sha256":"8f2d8ba2c309f8bbf227b18a6caa403552fbbeb164668f8e08daaba0c8cd4e91"},"schema_version":"1.0","source":{"id":"2501.06638","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2501.06638","created_at":"2026-07-05T10:00:07Z"},{"alias_kind":"arxiv_version","alias_value":"2501.06638v1","created_at":"2026-07-05T10:00:07Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.06638","created_at":"2026-07-05T10:00:07Z"},{"alias_kind":"pith_short_12","alias_value":"3J7SZEDFLDVD","created_at":"2026-07-05T10:00:07Z"},{"alias_kind":"pith_short_16","alias_value":"3J7SZEDFLDVDWUSO","created_at":"2026-07-05T10:00:07Z"},{"alias_kind":"pith_short_8","alias_value":"3J7SZEDF","created_at":"2026-07-05T10:00:07Z"}],"graph_snapshots":[{"event_id":"sha256:bafb3d2bdb0345d8331e7c6781f6332587b2c6fffc57cf72f70fa23c4a8dbee2","target":"graph","created_at":"2026-07-05T10:00:07Z","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/2501.06638/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Semantic leakage is a phenomenon recently introduced by Gonen et al. (2024). It refers to a situation in which associations learnt from the training data emerge in language model generations in an unexpected and sometimes undesired way. Prior work has focused on leakage in large language models (7B+ parameters). In this study, I use Qwen2.5 model family to explore whether smaller models, ranging from 500M to 7B parameters, demonstrate less semantic leakage due to their limited capacity for capturing complex associations. Building on the previous dataset from Gonen et al. (2024), I introduce a ","authors_text":"Veronika Smilga","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-11T21:03:22Z","title":"Scaling Down Semantic Leakage: Investigating Associative Bias in Smaller Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.06638","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:ef4a605591a8be8b4fa5f215189aadd631b64cf09b14d4b96b2ea44b617aa4fe","target":"record","created_at":"2026-07-05T10:00:07Z","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":"ea76cd5387087a1763abfb99bb0b654e834db41f65eb7ec9a05750b9e1bb9c41","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-11T21:03:22Z","title_canon_sha256":"8f2d8ba2c309f8bbf227b18a6caa403552fbbeb164668f8e08daaba0c8cd4e91"},"schema_version":"1.0","source":{"id":"2501.06638","kind":"arxiv","version":1}},"canonical_sha256":"da7f2c906558ea3b524ef570611b4bc20691eddc24b03b0a828d3f174edc3be5","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"da7f2c906558ea3b524ef570611b4bc20691eddc24b03b0a828d3f174edc3be5","first_computed_at":"2026-07-05T10:00:07.630966Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:00:07.630966Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"DN5ajRQyOB3iFnMLx8q5wqfqpZqkj1x1+HNJMoO/A+BiH9G6cxfZgohQ8GlfN03qU4bRIpzVT/atDF30hKYMCA==","signature_status":"signed_v1","signed_at":"2026-07-05T10:00:07.631403Z","signed_message":"canonical_sha256_bytes"},"source_id":"2501.06638","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:ef4a605591a8be8b4fa5f215189aadd631b64cf09b14d4b96b2ea44b617aa4fe","sha256:bafb3d2bdb0345d8331e7c6781f6332587b2c6fffc57cf72f70fa23c4a8dbee2"],"state_sha256":"6f2dd7d4086017779d906367db19a22693cd4ef465dfc21b32cfceedca4f6481"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"1IKtXmREyVjFkobn8CGszx/0o/wiJaSmEFolUqG6IzzmMPJSWgf1v/+1IFmsnRFTOyCiM4UMvLUF7Iz4dAa5Ag==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-11T19:41:50.016899Z","bundle_sha256":"92921090f598cd1d3eeeb0e37ebb73513ba0817d7758c1461c0810d0bb701d15"}}