{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:MCKA6H4H2AIAJPSHZLQFCJRSP2","short_pith_number":"pith:MCKA6H4H","canonical_record":{"source":{"id":"2601.00417","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-01-01T18:11:38Z","cross_cats_sorted":["cs.AI","cs.CL","cs.CV"],"title_canon_sha256":"7b8a6309ac3992381c549c47e19a9b881f02eec4e1a89f46cfda0566c3dba6ab","abstract_canon_sha256":"875e5e10e3508eb1b9efb7e96a35e54a6142a36079e6a9bfac28b6f18446df04"},"schema_version":"1.0"},"canonical_sha256":"60940f1f87d01004be47cae05126327ea06b8e6308ba65963df8d7e917432728","source":{"kind":"arxiv","id":"2601.00417","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2601.00417","created_at":"2026-05-18T03:10:11Z"},{"alias_kind":"arxiv_version","alias_value":"2601.00417v3","created_at":"2026-05-18T03:10:11Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2601.00417","created_at":"2026-05-18T03:10:11Z"},{"alias_kind":"pith_short_12","alias_value":"MCKA6H4H2AIA","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"MCKA6H4H2AIAJPSH","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"MCKA6H4H","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:MCKA6H4H2AIAJPSHZLQFCJRSP2","target":"record","payload":{"canonical_record":{"source":{"id":"2601.00417","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-01-01T18:11:38Z","cross_cats_sorted":["cs.AI","cs.CL","cs.CV"],"title_canon_sha256":"7b8a6309ac3992381c549c47e19a9b881f02eec4e1a89f46cfda0566c3dba6ab","abstract_canon_sha256":"875e5e10e3508eb1b9efb7e96a35e54a6142a36079e6a9bfac28b6f18446df04"},"schema_version":"1.0"},"canonical_sha256":"60940f1f87d01004be47cae05126327ea06b8e6308ba65963df8d7e917432728","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T03:10:11.457329Z","signature_b64":"0V5hhR4xRVZOgpRDV8F/Ig6UC+j/S+429wI++Aau218exbiE7PseNIrKEQAoAJOoATof1TCXjt49B+six26pAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"60940f1f87d01004be47cae05126327ea06b8e6308ba65963df8d7e917432728","last_reissued_at":"2026-05-18T03:10:11.456646Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T03:10:11.456646Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2601.00417","source_version":3,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-18T03:10:11Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"TXjOWEh8cfebdfrTHzZSbszjHLs+wOd9tr31pk2aWgWeNCIhl42UIArv5/OU86rHQe1ahPvxCjU8rf1FYfClCw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-23T09:41:48.898929Z"},"content_sha256":"011c5ba3beb944717a96dc76cf79a60b25c0e325eef4d9c3d147a2e954e7b90d","schema_version":"1.0","event_id":"sha256:011c5ba3beb944717a96dc76cf79a60b25c0e325eef4d9c3d147a2e954e7b90d"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:MCKA6H4H2AIAJPSHZLQFCJRSP2","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Deep Delta Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Deep Delta Learning lets Transformer layers selectively rewrite residual content instead of only adding to it.","cross_cats":["cs.AI","cs.CL","cs.CV"],"primary_cat":"cs.LG","authors_text":"Mengdi Wang, Quanquan Gu, Yifan Zhang, Yifeng Liu","submitted_at":"2026-01-01T18:11:38Z","abstract_excerpt":"Transformer residual streams evolve by additive accumulation: each layer appends a feature update to a shared hidden state, but has no direct mechanism for replacing content that has become obsolete or conflicting. We introduce Deep Delta Learning (DDL), a residual update rule that preserves the identity path while giving every layer the ability to selectively rewrite residual content. DDL reads the current state along a learned direction, compares it with a learned target value, and writes back a gated correction along the same direction. When the gate is closed, the update reduces to the ide"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Controlled pretraining and downstream evaluations show that residual rewrite operations improve language modeling quality relative to pure additive accumulation introduced in ResNet.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the learned directions, target values, and gates can reliably identify and correct obsolete or conflicting residual content without introducing training instability or degrading the identity path.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Deep Delta Learning replaces additive residual updates with a gated delta-rule that selectively overwrites residual content along learned directions, improving language modeling quality over standard ResNet-style accumulation.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Deep Delta Learning lets Transformer layers selectively rewrite residual content instead of only adding to it.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"376c439b656452e84b41d3291c318c6f3811a0b4a2ea8e3ce1a222362a125599"},"source":{"id":"2601.00417","kind":"arxiv","version":3},"verdict":{"id":"7ab30361-8111-4ea2-b250-9cd4f6a0d945","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T17:43:49.467892Z","strongest_claim":"Controlled pretraining and downstream evaluations show that residual rewrite operations improve language modeling quality relative to pure additive accumulation introduced in ResNet.","one_line_summary":"Deep Delta Learning replaces additive residual updates with a gated delta-rule that selectively overwrites residual content along learned directions, improving language modeling quality over standard ResNet-style accumulation.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the learned directions, target values, and gates can reliably identify and correct obsolete or conflicting residual content without introducing training instability or degrading the identity path.","pith_extraction_headline":"Deep Delta Learning lets Transformer layers selectively rewrite residual content instead of only adding to it."},"references":{"count":20,"sample":[{"doi":"","year":null,"title":"Hoft: Householder orthogonal fine-tuning.arXiv preprint arXiv:2505.16531,","work_id":"143c72d8-47ea-4125-9fb2-fff1c9f96381","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2010,"title":"N-ode transformer: A depth-adaptive variant of the transformer using neural ordinary differential equations.arXiv preprint arXiv:2010.11358,","work_id":"3408669e-3173-4a4c-8509-8524398d3587","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1905,"title":"BoolQ: Exploring the Surprising Difficulty of Natural Yes/No Questions","work_id":"511eeb84-4b95-46d5-b14f-50da43f4f19f","ref_index":3,"cited_arxiv_id":"1905.10044","is_internal_anchor":true},{"doi":"","year":null,"title":"arXiv preprint arXiv:2201.12133 , year=","work_id":"3370ebbd-0dc5-4946-b272-dca02a58236e","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Chaos meets attention: Transformers for large-scale dynamical prediction.arXiv preprint arXiv:2504.20858,","work_id":"b531d814-bd06-444a-94f5-92b0a25ce42b","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":20,"snapshot_sha256":"bf176bc54c4aff10a6e8eae9127305668dbb3feac7f8fbb0cdae701f2332b1d5","internal_anchors":7},"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":"7ab30361-8111-4ea2-b250-9cd4f6a0d945"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-18T03:10:11Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"dnazd8dw5CR4Mvo4vG5iRzqB68NR6Dkn8EHsNWtxdIbh8Zx648Hj7kMFB5mcpIVKi+eo2L7LcOKq0TL+QkMRCA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-23T09:41:48.900026Z"},"content_sha256":"8910dfea2028f332087160c3ad64a158a1df45409fbf31991f3b9a8c42efa169","schema_version":"1.0","event_id":"sha256:8910dfea2028f332087160c3ad64a158a1df45409fbf31991f3b9a8c42efa169"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/MCKA6H4H2AIAJPSHZLQFCJRSP2/bundle.json","state_url":"https://pith.science/pith/MCKA6H4H2AIAJPSHZLQFCJRSP2/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/MCKA6H4H2AIAJPSHZLQFCJRSP2/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":"2026-05-23T09:41:48Z","links":{"resolver":"https://pith.science/pith/MCKA6H4H2AIAJPSHZLQFCJRSP2","bundle":"https://pith.science/pith/MCKA6H4H2AIAJPSHZLQFCJRSP2/bundle.json","state":"https://pith.science/pith/MCKA6H4H2AIAJPSHZLQFCJRSP2/state.json","well_known_bundle":"https://pith.science/.well-known/pith/MCKA6H4H2AIAJPSHZLQFCJRSP2/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:MCKA6H4H2AIAJPSHZLQFCJRSP2","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":"875e5e10e3508eb1b9efb7e96a35e54a6142a36079e6a9bfac28b6f18446df04","cross_cats_sorted":["cs.AI","cs.CL","cs.CV"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-01-01T18:11:38Z","title_canon_sha256":"7b8a6309ac3992381c549c47e19a9b881f02eec4e1a89f46cfda0566c3dba6ab"},"schema_version":"1.0","source":{"id":"2601.00417","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2601.00417","created_at":"2026-05-18T03:10:11Z"},{"alias_kind":"arxiv_version","alias_value":"2601.00417v3","created_at":"2026-05-18T03:10:11Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2601.00417","created_at":"2026-05-18T03:10:11Z"},{"alias_kind":"pith_short_12","alias_value":"MCKA6H4H2AIA","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"MCKA6H4H2AIAJPSH","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"MCKA6H4H","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:8910dfea2028f332087160c3ad64a158a1df45409fbf31991f3b9a8c42efa169","target":"graph","created_at":"2026-05-18T03:10:11Z","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":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Controlled pretraining and downstream evaluations show that residual rewrite operations improve language modeling quality relative to pure additive accumulation introduced in ResNet."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the learned directions, target values, and gates can reliably identify and correct obsolete or conflicting residual content without introducing training instability or degrading the identity path."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Deep Delta Learning replaces additive residual updates with a gated delta-rule that selectively overwrites residual content along learned directions, improving language modeling quality over standard ResNet-style accumulation."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Deep Delta Learning lets Transformer layers selectively rewrite residual content instead of only adding to it."}],"snapshot_sha256":"376c439b656452e84b41d3291c318c6f3811a0b4a2ea8e3ce1a222362a125599"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"paper":{"abstract_excerpt":"Transformer residual streams evolve by additive accumulation: each layer appends a feature update to a shared hidden state, but has no direct mechanism for replacing content that has become obsolete or conflicting. We introduce Deep Delta Learning (DDL), a residual update rule that preserves the identity path while giving every layer the ability to selectively rewrite residual content. DDL reads the current state along a learned direction, compares it with a learned target value, and writes back a gated correction along the same direction. When the gate is closed, the update reduces to the ide","authors_text":"Mengdi Wang, Quanquan Gu, Yifan Zhang, Yifeng Liu","cross_cats":["cs.AI","cs.CL","cs.CV"],"headline":"Deep Delta Learning lets Transformer layers selectively rewrite residual content instead of only adding to it.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-01-01T18:11:38Z","title":"Deep Delta Learning"},"references":{"count":20,"internal_anchors":7,"resolved_work":20,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Hoft: Householder orthogonal fine-tuning.arXiv preprint arXiv:2505.16531,","work_id":"143c72d8-47ea-4125-9fb2-fff1c9f96381","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"N-ode transformer: A depth-adaptive variant of the transformer using neural ordinary differential equations.arXiv preprint arXiv:2010.11358,","work_id":"3408669e-3173-4a4c-8509-8524398d3587","year":2010},{"cited_arxiv_id":"1905.10044","doi":"","is_internal_anchor":true,"ref_index":3,"title":"BoolQ: Exploring the Surprising Difficulty of Natural Yes/No Questions","work_id":"511eeb84-4b95-46d5-b14f-50da43f4f19f","year":1905},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"arXiv preprint arXiv:2201.12133 , year=","work_id":"3370ebbd-0dc5-4946-b272-dca02a58236e","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Chaos meets attention: Transformers for large-scale dynamical prediction.arXiv preprint arXiv:2504.20858,","work_id":"b531d814-bd06-444a-94f5-92b0a25ce42b","year":null}],"snapshot_sha256":"bf176bc54c4aff10a6e8eae9127305668dbb3feac7f8fbb0cdae701f2332b1d5"},"source":{"id":"2601.00417","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-16T17:43:49.467892Z","id":"7ab30361-8111-4ea2-b250-9cd4f6a0d945","model_set":{"reader":"grok-4.3"},"one_line_summary":"Deep Delta Learning replaces additive residual updates with a gated delta-rule that selectively overwrites residual content along learned directions, improving language modeling quality over standard ResNet-style accumulation.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Deep Delta Learning lets Transformer layers selectively rewrite residual content instead of only adding to it.","strongest_claim":"Controlled pretraining and downstream evaluations show that residual rewrite operations improve language modeling quality relative to pure additive accumulation introduced in ResNet.","weakest_assumption":"That the learned directions, target values, and gates can reliably identify and correct obsolete or conflicting residual content without introducing training instability or degrading the identity path."}},"verdict_id":"7ab30361-8111-4ea2-b250-9cd4f6a0d945"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:011c5ba3beb944717a96dc76cf79a60b25c0e325eef4d9c3d147a2e954e7b90d","target":"record","created_at":"2026-05-18T03:10:11Z","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":"875e5e10e3508eb1b9efb7e96a35e54a6142a36079e6a9bfac28b6f18446df04","cross_cats_sorted":["cs.AI","cs.CL","cs.CV"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-01-01T18:11:38Z","title_canon_sha256":"7b8a6309ac3992381c549c47e19a9b881f02eec4e1a89f46cfda0566c3dba6ab"},"schema_version":"1.0","source":{"id":"2601.00417","kind":"arxiv","version":3}},"canonical_sha256":"60940f1f87d01004be47cae05126327ea06b8e6308ba65963df8d7e917432728","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"60940f1f87d01004be47cae05126327ea06b8e6308ba65963df8d7e917432728","first_computed_at":"2026-05-18T03:10:11.456646Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-18T03:10:11.456646Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"0V5hhR4xRVZOgpRDV8F/Ig6UC+j/S+429wI++Aau218exbiE7PseNIrKEQAoAJOoATof1TCXjt49B+six26pAw==","signature_status":"signed_v1","signed_at":"2026-05-18T03:10:11.457329Z","signed_message":"canonical_sha256_bytes"},"source_id":"2601.00417","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:011c5ba3beb944717a96dc76cf79a60b25c0e325eef4d9c3d147a2e954e7b90d","sha256:8910dfea2028f332087160c3ad64a158a1df45409fbf31991f3b9a8c42efa169"],"state_sha256":"8ee284296235a6f73e7d5a4b113a1e1a7d03b17a128658c12687f6ce57a7d803"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"qxut2yt+fPIT5Clk5bIW5fXVr7zTRlRTTDoqBjTWjK4k4oQKTz7Sbh4Csj80iMvFp0XtpczdqkmVH3KnnWL+BQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-23T09:41:48.905303Z","bundle_sha256":"84ca4bc1e2e6ddc3c78664e1d027be07cf171dbb5098f5319912e8b2c051521e"}}