{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2021:2CNWIVJV6RBUEBUEUMSMVLTIW4","short_pith_number":"pith:2CNWIVJV","canonical_record":{"source":{"id":"2104.06129","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-04-13T12:04:30Z","cross_cats_sorted":[],"title_canon_sha256":"b52fa12ea63774ef743c1cf4a7a0e05c5383b08f1a8d22d5349a54bd1f379462","abstract_canon_sha256":"efaf596fcc97f2b328b51bee91a451280597dc3b6dcde96f84418cbbd42b60b9"},"schema_version":"1.0"},"canonical_sha256":"d09b645535f443420684a324caae68b7025e3b6a7caf65fe49a77878834805c3","source":{"kind":"arxiv","id":"2104.06129","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2104.06129","created_at":"2026-07-05T03:11:28Z"},{"alias_kind":"arxiv_version","alias_value":"2104.06129v2","created_at":"2026-07-05T03:11:28Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2104.06129","created_at":"2026-07-05T03:11:28Z"},{"alias_kind":"pith_short_12","alias_value":"2CNWIVJV6RBU","created_at":"2026-07-05T03:11:28Z"},{"alias_kind":"pith_short_16","alias_value":"2CNWIVJV6RBUEBUE","created_at":"2026-07-05T03:11:28Z"},{"alias_kind":"pith_short_8","alias_value":"2CNWIVJV","created_at":"2026-07-05T03:11:28Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2021:2CNWIVJV6RBUEBUEUMSMVLTIW4","target":"record","payload":{"canonical_record":{"source":{"id":"2104.06129","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-04-13T12:04:30Z","cross_cats_sorted":[],"title_canon_sha256":"b52fa12ea63774ef743c1cf4a7a0e05c5383b08f1a8d22d5349a54bd1f379462","abstract_canon_sha256":"efaf596fcc97f2b328b51bee91a451280597dc3b6dcde96f84418cbbd42b60b9"},"schema_version":"1.0"},"canonical_sha256":"d09b645535f443420684a324caae68b7025e3b6a7caf65fe49a77878834805c3","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:11:28.914540Z","signature_b64":"ZRujS4g6jMyoV82KBPc/96W8Alw2Dv0UeNKtDCGy3vdCm7Qt+1+Cseimx0/6tf7aRy9/22L2agzdDnKWMFLrAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d09b645535f443420684a324caae68b7025e3b6a7caf65fe49a77878834805c3","last_reissued_at":"2026-07-05T03:11:28.914091Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:11:28.914091Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2104.06129","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":"2026-07-05T03:11:28Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"TjGB8q5qQYhIFvIkVdfkHHxnPGRg7yEGF33t7Pw8I3QMdBe/AMX4yaEbQFVqP4iraN52mA9FzAg3dGIMfCpVCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-16T04:00:04.742859Z"},"content_sha256":"72aab4ba6ba9f15e058b71ee61c0a15c4314961faa0b5df12070e52ec64232cd","schema_version":"1.0","event_id":"sha256:72aab4ba6ba9f15e058b71ee61c0a15c4314961faa0b5df12070e52ec64232cd"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2021:2CNWIVJV6RBUEBUEUMSMVLTIW4","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"What's in your Head? Emergent Behaviour in Multi-Task Transformer Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aviv Ben-Arie, Jonathan Berant, Mor Geva, Uri Katz","submitted_at":"2021-04-13T12:04:30Z","abstract_excerpt":"The primary paradigm for multi-task training in natural language processing is to represent the input with a shared pre-trained language model, and add a small, thin network (head) per task. Given an input, a target head is the head that is selected for outputting the final prediction. In this work, we examine the behaviour of non-target heads, that is, the output of heads when given input that belongs to a different task than the one they were trained for. We find that non-target heads exhibit emergent behaviour, which may either explain the target task, or generalize beyond their original ta"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2104.06129","kind":"arxiv","version":2},"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/2104.06129/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":"2026-07-05T03:11:28Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"qARPd+Cbkg2NgoQb9CZinOf1eG+XYKJrODfzDAM83yg8RjPBELbW2btlTC1w9jCwZJf3lrd2hRo1jYRuVXQ/Cg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-16T04:00:04.743237Z"},"content_sha256":"2fd32bd78316128dcb1f40c566ff9c82ff6e01af88dc82b5b8b9cc8c8b055395","schema_version":"1.0","event_id":"sha256:2fd32bd78316128dcb1f40c566ff9c82ff6e01af88dc82b5b8b9cc8c8b055395"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/2CNWIVJV6RBUEBUEUMSMVLTIW4/bundle.json","state_url":"https://pith.science/pith/2CNWIVJV6RBUEBUEUMSMVLTIW4/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/2CNWIVJV6RBUEBUEUMSMVLTIW4/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":"2026-07-16T04:00:04Z","links":{"resolver":"https://pith.science/pith/2CNWIVJV6RBUEBUEUMSMVLTIW4","bundle":"https://pith.science/pith/2CNWIVJV6RBUEBUEUMSMVLTIW4/bundle.json","state":"https://pith.science/pith/2CNWIVJV6RBUEBUEUMSMVLTIW4/state.json","well_known_bundle":"https://pith.science/.well-known/pith/2CNWIVJV6RBUEBUEUMSMVLTIW4/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2021:2CNWIVJV6RBUEBUEUMSMVLTIW4","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":"efaf596fcc97f2b328b51bee91a451280597dc3b6dcde96f84418cbbd42b60b9","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-04-13T12:04:30Z","title_canon_sha256":"b52fa12ea63774ef743c1cf4a7a0e05c5383b08f1a8d22d5349a54bd1f379462"},"schema_version":"1.0","source":{"id":"2104.06129","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2104.06129","created_at":"2026-07-05T03:11:28Z"},{"alias_kind":"arxiv_version","alias_value":"2104.06129v2","created_at":"2026-07-05T03:11:28Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2104.06129","created_at":"2026-07-05T03:11:28Z"},{"alias_kind":"pith_short_12","alias_value":"2CNWIVJV6RBU","created_at":"2026-07-05T03:11:28Z"},{"alias_kind":"pith_short_16","alias_value":"2CNWIVJV6RBUEBUE","created_at":"2026-07-05T03:11:28Z"},{"alias_kind":"pith_short_8","alias_value":"2CNWIVJV","created_at":"2026-07-05T03:11:28Z"}],"graph_snapshots":[{"event_id":"sha256:2fd32bd78316128dcb1f40c566ff9c82ff6e01af88dc82b5b8b9cc8c8b055395","target":"graph","created_at":"2026-07-05T03:11:28Z","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/2104.06129/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"The primary paradigm for multi-task training in natural language processing is to represent the input with a shared pre-trained language model, and add a small, thin network (head) per task. Given an input, a target head is the head that is selected for outputting the final prediction. In this work, we examine the behaviour of non-target heads, that is, the output of heads when given input that belongs to a different task than the one they were trained for. We find that non-target heads exhibit emergent behaviour, which may either explain the target task, or generalize beyond their original ta","authors_text":"Aviv Ben-Arie, Jonathan Berant, Mor Geva, Uri Katz","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-04-13T12:04:30Z","title":"What's in your Head? Emergent Behaviour in Multi-Task Transformer Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2104.06129","kind":"arxiv","version":2},"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":"sha256:72aab4ba6ba9f15e058b71ee61c0a15c4314961faa0b5df12070e52ec64232cd","target":"record","created_at":"2026-07-05T03:11:28Z","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":"efaf596fcc97f2b328b51bee91a451280597dc3b6dcde96f84418cbbd42b60b9","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-04-13T12:04:30Z","title_canon_sha256":"b52fa12ea63774ef743c1cf4a7a0e05c5383b08f1a8d22d5349a54bd1f379462"},"schema_version":"1.0","source":{"id":"2104.06129","kind":"arxiv","version":2}},"canonical_sha256":"d09b645535f443420684a324caae68b7025e3b6a7caf65fe49a77878834805c3","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"d09b645535f443420684a324caae68b7025e3b6a7caf65fe49a77878834805c3","first_computed_at":"2026-07-05T03:11:28.914091Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T03:11:28.914091Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"ZRujS4g6jMyoV82KBPc/96W8Alw2Dv0UeNKtDCGy3vdCm7Qt+1+Cseimx0/6tf7aRy9/22L2agzdDnKWMFLrAw==","signature_status":"signed_v1","signed_at":"2026-07-05T03:11:28.914540Z","signed_message":"canonical_sha256_bytes"},"source_id":"2104.06129","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:72aab4ba6ba9f15e058b71ee61c0a15c4314961faa0b5df12070e52ec64232cd","sha256:2fd32bd78316128dcb1f40c566ff9c82ff6e01af88dc82b5b8b9cc8c8b055395"],"state_sha256":"8658e971fd07e997a37c7f0862067863fe4675789105978f05215d0d9ac61a86"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"yKGOjMX5ykS+kjVxz1MnH1EaDfXScYSbSLWPjevzA2td29yaNEZNJfUGL0gPZzAxr5GZRaOvAPa9cPVoSAgXBA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-16T04:00:04.745430Z","bundle_sha256":"ee5645a891e67d12aa1956506678b33e39e1dfa46b9d96fb59e79e56aa7d9baa"}}