{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2020:TYMWKV4ZH5Y3YKHIW6YFG2IXSJ","short_pith_number":"pith:TYMWKV4Z","canonical_record":{"source":{"id":"2004.04487","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-04-09T11:14:56Z","cross_cats_sorted":[],"title_canon_sha256":"35bdaa53d5171edab504685568a65aefd70524e213dcb5f89f1d1a46e21ee03e","abstract_canon_sha256":"1107f49f41bcee6874e5ff41e59cf79d21c48a211bfab6378cb97d143fcccd7c"},"schema_version":"1.0"},"canonical_sha256":"9e196557993f71bc28e8b7b0536917924b08b9fb9a04c8563635a7290da3ef77","source":{"kind":"arxiv","id":"2004.04487","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2004.04487","created_at":"2026-07-05T00:54:08Z"},{"alias_kind":"arxiv_version","alias_value":"2004.04487v1","created_at":"2026-07-05T00:54:08Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2004.04487","created_at":"2026-07-05T00:54:08Z"},{"alias_kind":"pith_short_12","alias_value":"TYMWKV4ZH5Y3","created_at":"2026-07-05T00:54:08Z"},{"alias_kind":"pith_short_16","alias_value":"TYMWKV4ZH5Y3YKHI","created_at":"2026-07-05T00:54:08Z"},{"alias_kind":"pith_short_8","alias_value":"TYMWKV4Z","created_at":"2026-07-05T00:54:08Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2020:TYMWKV4ZH5Y3YKHIW6YFG2IXSJ","target":"record","payload":{"canonical_record":{"source":{"id":"2004.04487","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-04-09T11:14:56Z","cross_cats_sorted":[],"title_canon_sha256":"35bdaa53d5171edab504685568a65aefd70524e213dcb5f89f1d1a46e21ee03e","abstract_canon_sha256":"1107f49f41bcee6874e5ff41e59cf79d21c48a211bfab6378cb97d143fcccd7c"},"schema_version":"1.0"},"canonical_sha256":"9e196557993f71bc28e8b7b0536917924b08b9fb9a04c8563635a7290da3ef77","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T00:54:08.559050Z","signature_b64":"7gfEeP2DgA5xShK54ljbp3+GyUaPJPYg1oV5lHqyUAcnMTS+3ygh17MyjtStaEw/S+iS/AitarsjQX9I3gg7Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9e196557993f71bc28e8b7b0536917924b08b9fb9a04c8563635a7290da3ef77","last_reissued_at":"2026-07-05T00:54:08.558708Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T00:54:08.558708Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2004.04487","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":"2026-07-05T00:54:08Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"d8WyJdweNDe4NJhOLYTqJs9djKFfeX28VGNvSoQbpte8p0D4/1rXLHBmjVwPFzzh6cNkJoEfqcMldZJUszjpBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T15:43:44.947674Z"},"content_sha256":"5072b7658b87dfbbb24a112e806731b92bc609312def334292e5c1d69558e81a","schema_version":"1.0","event_id":"sha256:5072b7658b87dfbbb24a112e806731b92bc609312def334292e5c1d69558e81a"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2020:TYMWKV4ZH5Y3YKHIW6YFG2IXSJ","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Injecting Numerical Reasoning Skills into Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Ankit Gupta, Jonathan Berant, Mor Geva","submitted_at":"2020-04-09T11:14:56Z","abstract_excerpt":"Large pre-trained language models (LMs) are known to encode substantial amounts of linguistic information. However, high-level reasoning skills, such as numerical reasoning, are difficult to learn from a language-modeling objective only. Consequently, existing models for numerical reasoning have used specialized architectures with limited flexibility. In this work, we show that numerical reasoning is amenable to automatic data generation, and thus one can inject this skill into pre-trained LMs, by generating large amounts of data, and training in a multi-task setup. We show that pre-training o"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2004.04487","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/2004.04487/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":"2026-07-05T00:54:08Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"/Sq9NJoqq8+IPeJCph31YPdHdLQX5vFhGbfZEKhRq6W/lE4U5lN+JtMnyM/+3FQiTC+miEiAIg3S7h+fi+SeBA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T15:43:44.948447Z"},"content_sha256":"b250d37e5f40e9ed99b4a866dcd6fd1c28bb3cfc63a2dadae4809f200ed176db","schema_version":"1.0","event_id":"sha256:b250d37e5f40e9ed99b4a866dcd6fd1c28bb3cfc63a2dadae4809f200ed176db"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/TYMWKV4ZH5Y3YKHIW6YFG2IXSJ/bundle.json","state_url":"https://pith.science/pith/TYMWKV4ZH5Y3YKHIW6YFG2IXSJ/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/TYMWKV4ZH5Y3YKHIW6YFG2IXSJ/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":"2026-08-04T15:43:44Z","links":{"resolver":"https://pith.science/pith/TYMWKV4ZH5Y3YKHIW6YFG2IXSJ","bundle":"https://pith.science/pith/TYMWKV4ZH5Y3YKHIW6YFG2IXSJ/bundle.json","state":"https://pith.science/pith/TYMWKV4ZH5Y3YKHIW6YFG2IXSJ/state.json","well_known_bundle":"https://pith.science/.well-known/pith/TYMWKV4ZH5Y3YKHIW6YFG2IXSJ/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2020:TYMWKV4ZH5Y3YKHIW6YFG2IXSJ","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":"1107f49f41bcee6874e5ff41e59cf79d21c48a211bfab6378cb97d143fcccd7c","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-04-09T11:14:56Z","title_canon_sha256":"35bdaa53d5171edab504685568a65aefd70524e213dcb5f89f1d1a46e21ee03e"},"schema_version":"1.0","source":{"id":"2004.04487","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2004.04487","created_at":"2026-07-05T00:54:08Z"},{"alias_kind":"arxiv_version","alias_value":"2004.04487v1","created_at":"2026-07-05T00:54:08Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2004.04487","created_at":"2026-07-05T00:54:08Z"},{"alias_kind":"pith_short_12","alias_value":"TYMWKV4ZH5Y3","created_at":"2026-07-05T00:54:08Z"},{"alias_kind":"pith_short_16","alias_value":"TYMWKV4ZH5Y3YKHI","created_at":"2026-07-05T00:54:08Z"},{"alias_kind":"pith_short_8","alias_value":"TYMWKV4Z","created_at":"2026-07-05T00:54:08Z"}],"graph_snapshots":[{"event_id":"sha256:b250d37e5f40e9ed99b4a866dcd6fd1c28bb3cfc63a2dadae4809f200ed176db","target":"graph","created_at":"2026-07-05T00:54:08Z","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/2004.04487/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Large pre-trained language models (LMs) are known to encode substantial amounts of linguistic information. However, high-level reasoning skills, such as numerical reasoning, are difficult to learn from a language-modeling objective only. Consequently, existing models for numerical reasoning have used specialized architectures with limited flexibility. In this work, we show that numerical reasoning is amenable to automatic data generation, and thus one can inject this skill into pre-trained LMs, by generating large amounts of data, and training in a multi-task setup. We show that pre-training o","authors_text":"Ankit Gupta, Jonathan Berant, Mor Geva","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-04-09T11:14:56Z","title":"Injecting Numerical Reasoning Skills into Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2004.04487","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":"sha256:5072b7658b87dfbbb24a112e806731b92bc609312def334292e5c1d69558e81a","target":"record","created_at":"2026-07-05T00:54:08Z","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":"1107f49f41bcee6874e5ff41e59cf79d21c48a211bfab6378cb97d143fcccd7c","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-04-09T11:14:56Z","title_canon_sha256":"35bdaa53d5171edab504685568a65aefd70524e213dcb5f89f1d1a46e21ee03e"},"schema_version":"1.0","source":{"id":"2004.04487","kind":"arxiv","version":1}},"canonical_sha256":"9e196557993f71bc28e8b7b0536917924b08b9fb9a04c8563635a7290da3ef77","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"9e196557993f71bc28e8b7b0536917924b08b9fb9a04c8563635a7290da3ef77","first_computed_at":"2026-07-05T00:54:08.558708Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T00:54:08.558708Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"7gfEeP2DgA5xShK54ljbp3+GyUaPJPYg1oV5lHqyUAcnMTS+3ygh17MyjtStaEw/S+iS/AitarsjQX9I3gg7Ag==","signature_status":"signed_v1","signed_at":"2026-07-05T00:54:08.559050Z","signed_message":"canonical_sha256_bytes"},"source_id":"2004.04487","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:5072b7658b87dfbbb24a112e806731b92bc609312def334292e5c1d69558e81a","sha256:b250d37e5f40e9ed99b4a866dcd6fd1c28bb3cfc63a2dadae4809f200ed176db"],"state_sha256":"b012fd535dbe63baee5471ff72ff5f2f12e9a4971df950042d0e68952ee0fd24"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"6swLHOGaoOQOfq4e7m3VRCJagPNxjOkwgZ0KlRgCIe9dqhzS/oIz8z6p+kWCej4+LwFXl+MvhHU1clDkLb5IDw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-04T15:43:44.957023Z","bundle_sha256":"3ed7a0f2075f965958279079b2edd6a0207d3183096061bb06b1c334d96e923a"}}