{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:WOJ6RXDR4ANB3EPQH6EW4MK7D7","short_pith_number":"pith:WOJ6RXDR","canonical_record":{"source":{"id":"2406.07358","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-06-11T15:26:57Z","cross_cats_sorted":["cs.CL","cs.CY","cs.LG"],"title_canon_sha256":"acd608ef5917620af19fbcc26f2d7d7fba7a8cbfd1bb0ff98944008b74f9cea8","abstract_canon_sha256":"35a2d5f19b0e3362958db79a4c43b729153a18427c77f3c8f8ef48c08c3f2337"},"schema_version":"1.0"},"canonical_sha256":"b393e8dc71e01a1d91f03f896e315f1ffb54e017a70f69dbb4958642c92a5147","source":{"kind":"arxiv","id":"2406.07358","version":4},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.07358","created_at":"2026-07-05T10:10:41Z"},{"alias_kind":"arxiv_version","alias_value":"2406.07358v4","created_at":"2026-07-05T10:10:41Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.07358","created_at":"2026-07-05T10:10:41Z"},{"alias_kind":"pith_short_12","alias_value":"WOJ6RXDR4ANB","created_at":"2026-07-05T10:10:41Z"},{"alias_kind":"pith_short_16","alias_value":"WOJ6RXDR4ANB3EPQ","created_at":"2026-07-05T10:10:41Z"},{"alias_kind":"pith_short_8","alias_value":"WOJ6RXDR","created_at":"2026-07-05T10:10:41Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:WOJ6RXDR4ANB3EPQH6EW4MK7D7","target":"record","payload":{"canonical_record":{"source":{"id":"2406.07358","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-06-11T15:26:57Z","cross_cats_sorted":["cs.CL","cs.CY","cs.LG"],"title_canon_sha256":"acd608ef5917620af19fbcc26f2d7d7fba7a8cbfd1bb0ff98944008b74f9cea8","abstract_canon_sha256":"35a2d5f19b0e3362958db79a4c43b729153a18427c77f3c8f8ef48c08c3f2337"},"schema_version":"1.0"},"canonical_sha256":"b393e8dc71e01a1d91f03f896e315f1ffb54e017a70f69dbb4958642c92a5147","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:10:41.147047Z","signature_b64":"JeZpW6QEdwNxEDmU71APVIrUGRTpjZifjxV3DzKSo0fk4xuwiejY0kfseSROUWO+kKL9ZiR9o61PtUSeNmdQCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b393e8dc71e01a1d91f03f896e315f1ffb54e017a70f69dbb4958642c92a5147","last_reissued_at":"2026-07-05T10:10:41.146617Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:10:41.146617Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2406.07358","source_version":4,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:10:41Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"XOaV3CxLD7zO1ibhE+530MuLq57OgKhrPe5y5buPzkEtUIyzBx5HZDWJhehKqRa3DXk4xr/dIviUwWQDq9jmBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-07T12:40:57.646420Z"},"content_sha256":"e84b19fbf7614cdc83189dae25b3a5256a73fa4c679b992994fcc4e1cea3e3f2","schema_version":"1.0","event_id":"sha256:e84b19fbf7614cdc83189dae25b3a5256a73fa4c679b992994fcc4e1cea3e3f2"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:WOJ6RXDR4ANB3EPQH6EW4MK7D7","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"AI Sandbagging: Language Models can Strategically Underperform on Evaluations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.CY","cs.LG"],"primary_cat":"cs.AI","authors_text":"Felix Hofst\\\"atter, Francis Rhys Ward, Ollie Jaffe, Samuel F. Brown, Teun van der Weij","submitted_at":"2024-06-11T15:26:57Z","abstract_excerpt":"Trustworthy capability evaluations are crucial for ensuring the safety of AI systems, and are becoming a key component of AI regulation. However, the developers of an AI system, or the AI system itself, may have incentives for evaluations to understate the AI's actual capability. These conflicting interests lead to the problem of sandbagging, which we define as strategic underperformance on an evaluation. In this paper we assess sandbagging capabilities in contemporary language models (LMs). We prompt frontier LMs, like GPT-4 and Claude 3 Opus, to selectively underperform on dangerous capabili"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.07358","kind":"arxiv","version":4},"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/2406.07358/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":"2026-07-05T10:10:41Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"wVi+XVYYb93SeuOG2a/qBEY/nE89J2lZ9SBQvHHXglh0Gtw0i21Aup4Vxp/tdejoCFRAmAGglCq1f77zwgnfBw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-07T12:40:57.646787Z"},"content_sha256":"8705e581a9e9ca8689dac2385a38454cfbb121b648ffb40d5d263ce295c7293c","schema_version":"1.0","event_id":"sha256:8705e581a9e9ca8689dac2385a38454cfbb121b648ffb40d5d263ce295c7293c"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/WOJ6RXDR4ANB3EPQH6EW4MK7D7/bundle.json","state_url":"https://pith.science/pith/WOJ6RXDR4ANB3EPQH6EW4MK7D7/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/WOJ6RXDR4ANB3EPQH6EW4MK7D7/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":"2026-07-07T12:40:57Z","links":{"resolver":"https://pith.science/pith/WOJ6RXDR4ANB3EPQH6EW4MK7D7","bundle":"https://pith.science/pith/WOJ6RXDR4ANB3EPQH6EW4MK7D7/bundle.json","state":"https://pith.science/pith/WOJ6RXDR4ANB3EPQH6EW4MK7D7/state.json","well_known_bundle":"https://pith.science/.well-known/pith/WOJ6RXDR4ANB3EPQH6EW4MK7D7/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:WOJ6RXDR4ANB3EPQH6EW4MK7D7","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":"35a2d5f19b0e3362958db79a4c43b729153a18427c77f3c8f8ef48c08c3f2337","cross_cats_sorted":["cs.CL","cs.CY","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-06-11T15:26:57Z","title_canon_sha256":"acd608ef5917620af19fbcc26f2d7d7fba7a8cbfd1bb0ff98944008b74f9cea8"},"schema_version":"1.0","source":{"id":"2406.07358","kind":"arxiv","version":4}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.07358","created_at":"2026-07-05T10:10:41Z"},{"alias_kind":"arxiv_version","alias_value":"2406.07358v4","created_at":"2026-07-05T10:10:41Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.07358","created_at":"2026-07-05T10:10:41Z"},{"alias_kind":"pith_short_12","alias_value":"WOJ6RXDR4ANB","created_at":"2026-07-05T10:10:41Z"},{"alias_kind":"pith_short_16","alias_value":"WOJ6RXDR4ANB3EPQ","created_at":"2026-07-05T10:10:41Z"},{"alias_kind":"pith_short_8","alias_value":"WOJ6RXDR","created_at":"2026-07-05T10:10:41Z"}],"graph_snapshots":[{"event_id":"sha256:8705e581a9e9ca8689dac2385a38454cfbb121b648ffb40d5d263ce295c7293c","target":"graph","created_at":"2026-07-05T10:10:41Z","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/2406.07358/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Trustworthy capability evaluations are crucial for ensuring the safety of AI systems, and are becoming a key component of AI regulation. However, the developers of an AI system, or the AI system itself, may have incentives for evaluations to understate the AI's actual capability. These conflicting interests lead to the problem of sandbagging, which we define as strategic underperformance on an evaluation. In this paper we assess sandbagging capabilities in contemporary language models (LMs). We prompt frontier LMs, like GPT-4 and Claude 3 Opus, to selectively underperform on dangerous capabili","authors_text":"Felix Hofst\\\"atter, Francis Rhys Ward, Ollie Jaffe, Samuel F. Brown, Teun van der Weij","cross_cats":["cs.CL","cs.CY","cs.LG"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-06-11T15:26:57Z","title":"AI Sandbagging: Language Models can Strategically Underperform on Evaluations"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.07358","kind":"arxiv","version":4},"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:e84b19fbf7614cdc83189dae25b3a5256a73fa4c679b992994fcc4e1cea3e3f2","target":"record","created_at":"2026-07-05T10:10:41Z","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":"35a2d5f19b0e3362958db79a4c43b729153a18427c77f3c8f8ef48c08c3f2337","cross_cats_sorted":["cs.CL","cs.CY","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-06-11T15:26:57Z","title_canon_sha256":"acd608ef5917620af19fbcc26f2d7d7fba7a8cbfd1bb0ff98944008b74f9cea8"},"schema_version":"1.0","source":{"id":"2406.07358","kind":"arxiv","version":4}},"canonical_sha256":"b393e8dc71e01a1d91f03f896e315f1ffb54e017a70f69dbb4958642c92a5147","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"b393e8dc71e01a1d91f03f896e315f1ffb54e017a70f69dbb4958642c92a5147","first_computed_at":"2026-07-05T10:10:41.146617Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:10:41.146617Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"JeZpW6QEdwNxEDmU71APVIrUGRTpjZifjxV3DzKSo0fk4xuwiejY0kfseSROUWO+kKL9ZiR9o61PtUSeNmdQCA==","signature_status":"signed_v1","signed_at":"2026-07-05T10:10:41.147047Z","signed_message":"canonical_sha256_bytes"},"source_id":"2406.07358","source_kind":"arxiv","source_version":4}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:e84b19fbf7614cdc83189dae25b3a5256a73fa4c679b992994fcc4e1cea3e3f2","sha256:8705e581a9e9ca8689dac2385a38454cfbb121b648ffb40d5d263ce295c7293c"],"state_sha256":"073325bda1263e8ff654b94b9192fd83d2ed78f17bb201a6ecbd234a9815e80c"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"dsiFZhbj+NZpfbeRgRf87/hdpkaGCcIdpQGB1jhlw98HjDDQOIbxZYs4JqSMGp3t8wjNcwvsNSHqVhlLq7PKCA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-07T12:40:57.648667Z","bundle_sha256":"be12df587ee8fe07ebdbdb3d5d88d2df0e277226731790ddd02eb541d1f572a4"}}