{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:IBP2FU7PMQQLLYZ53QLDTWK576","short_pith_number":"pith:IBP2FU7P","canonical_record":{"source":{"id":"2607.04640","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-07-06T03:51:24Z","cross_cats_sorted":[],"title_canon_sha256":"1e05dd9d6850bd2fefc6dbd6c9e658b71fbe964f25be8e557452f8037cb41906","abstract_canon_sha256":"baa9097be5e017aa2b4dfb923e44ccb3cd246a986c662d9f7069a84a942393d9"},"schema_version":"1.0"},"canonical_sha256":"405fa2d3ef6420b5e33ddc1639d95dffb7240f6db0efd5c1efa8de177ead74e3","source":{"kind":"arxiv","id":"2607.04640","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2607.04640","created_at":"2026-07-07T02:19:29Z"},{"alias_kind":"arxiv_version","alias_value":"2607.04640v1","created_at":"2026-07-07T02:19:29Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.04640","created_at":"2026-07-07T02:19:29Z"},{"alias_kind":"pith_short_12","alias_value":"IBP2FU7PMQQL","created_at":"2026-07-07T02:19:29Z"},{"alias_kind":"pith_short_16","alias_value":"IBP2FU7PMQQLLYZ5","created_at":"2026-07-07T02:19:29Z"},{"alias_kind":"pith_short_8","alias_value":"IBP2FU7P","created_at":"2026-07-07T02:19:29Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:IBP2FU7PMQQLLYZ53QLDTWK576","target":"record","payload":{"canonical_record":{"source":{"id":"2607.04640","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-07-06T03:51:24Z","cross_cats_sorted":[],"title_canon_sha256":"1e05dd9d6850bd2fefc6dbd6c9e658b71fbe964f25be8e557452f8037cb41906","abstract_canon_sha256":"baa9097be5e017aa2b4dfb923e44ccb3cd246a986c662d9f7069a84a942393d9"},"schema_version":"1.0"},"canonical_sha256":"405fa2d3ef6420b5e33ddc1639d95dffb7240f6db0efd5c1efa8de177ead74e3","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-07T02:19:29.046246Z","signature_b64":"z9nuUrdCIUAOFgSxfe+TlLPQS/dngKSwE29L0nUn8D/aTEy9iacunIALR2NcbtW4/qooYNEkAqRdDxvUnZnmAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"405fa2d3ef6420b5e33ddc1639d95dffb7240f6db0efd5c1efa8de177ead74e3","last_reissued_at":"2026-07-07T02:19:29.045572Z","signature_status":"signed_v1","first_computed_at":"2026-07-07T02:19:29.045572Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2607.04640","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":"2026-07-07T02:19:29Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"4ue1EYfSYqobm/Z+R7vGgIVmD8MK9YB+8l4l4VBkWHZBmOORhvy3NV8nh19IC+34xYgmWALLFO578ly1Di6NDA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-15T12:39:02.299421Z"},"content_sha256":"8ad1004e6ebe577d0e2f678cdc92ca413b9c21352b047810d4743d0a72942552","schema_version":"1.0","event_id":"sha256:8ad1004e6ebe577d0e2f678cdc92ca413b9c21352b047810d4743d0a72942552"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:IBP2FU7PMQQLLYZ53QLDTWK576","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Wrong Before Right: Late Rescue and Interface Failure in Aligned Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jiaqi Deng","submitted_at":"2026-07-06T03:51:24Z","abstract_excerpt":"We study how correctness is assembled inside aligned language models, not only whether the final answer is right. Using layer-wise difference-in-differences (DiD) trajectories over polarity-controlled minimal pairs, we identify the wrong-dip: in mid layers (25-90% depth), internal preference transiently commits to the incorrect answer and is rescued only by late-layer correction. We verify this causally with patchscope-style activation transplantation across 17 models, three families, and 64x scale (0.5B-32B). Four findings follow. (1) Alignment amplification of the causal wrong-dip is recipe-"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.04640","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/2607.04640/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":"2026-07-07T02:19:29Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"E31GAamEVqJVWsE03u4OiQYP2AjdE28NMl+jhz13JvDJ9DBCul7/AUlv6uYBsT5AgWMl8cND/LfkOpYlWNtyDA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-15T12:39:02.299796Z"},"content_sha256":"7f23484653fc44b1899997397d2fa7a46e81f4f1f53ff9cae1f96fa9fe23e755","schema_version":"1.0","event_id":"sha256:7f23484653fc44b1899997397d2fa7a46e81f4f1f53ff9cae1f96fa9fe23e755"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/IBP2FU7PMQQLLYZ53QLDTWK576/bundle.json","state_url":"https://pith.science/pith/IBP2FU7PMQQLLYZ53QLDTWK576/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/IBP2FU7PMQQLLYZ53QLDTWK576/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":"2026-07-15T12:39:02Z","links":{"resolver":"https://pith.science/pith/IBP2FU7PMQQLLYZ53QLDTWK576","bundle":"https://pith.science/pith/IBP2FU7PMQQLLYZ53QLDTWK576/bundle.json","state":"https://pith.science/pith/IBP2FU7PMQQLLYZ53QLDTWK576/state.json","well_known_bundle":"https://pith.science/.well-known/pith/IBP2FU7PMQQLLYZ53QLDTWK576/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:IBP2FU7PMQQLLYZ53QLDTWK576","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":"baa9097be5e017aa2b4dfb923e44ccb3cd246a986c662d9f7069a84a942393d9","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-07-06T03:51:24Z","title_canon_sha256":"1e05dd9d6850bd2fefc6dbd6c9e658b71fbe964f25be8e557452f8037cb41906"},"schema_version":"1.0","source":{"id":"2607.04640","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2607.04640","created_at":"2026-07-07T02:19:29Z"},{"alias_kind":"arxiv_version","alias_value":"2607.04640v1","created_at":"2026-07-07T02:19:29Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.04640","created_at":"2026-07-07T02:19:29Z"},{"alias_kind":"pith_short_12","alias_value":"IBP2FU7PMQQL","created_at":"2026-07-07T02:19:29Z"},{"alias_kind":"pith_short_16","alias_value":"IBP2FU7PMQQLLYZ5","created_at":"2026-07-07T02:19:29Z"},{"alias_kind":"pith_short_8","alias_value":"IBP2FU7P","created_at":"2026-07-07T02:19:29Z"}],"graph_snapshots":[{"event_id":"sha256:7f23484653fc44b1899997397d2fa7a46e81f4f1f53ff9cae1f96fa9fe23e755","target":"graph","created_at":"2026-07-07T02:19:29Z","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/2607.04640/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"We study how correctness is assembled inside aligned language models, not only whether the final answer is right. Using layer-wise difference-in-differences (DiD) trajectories over polarity-controlled minimal pairs, we identify the wrong-dip: in mid layers (25-90% depth), internal preference transiently commits to the incorrect answer and is rescued only by late-layer correction. We verify this causally with patchscope-style activation transplantation across 17 models, three families, and 64x scale (0.5B-32B). Four findings follow. (1) Alignment amplification of the causal wrong-dip is recipe-","authors_text":"Jiaqi Deng","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-07-06T03:51:24Z","title":"Wrong Before Right: Late Rescue and Interface Failure in Aligned Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.04640","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":"sha256:8ad1004e6ebe577d0e2f678cdc92ca413b9c21352b047810d4743d0a72942552","target":"record","created_at":"2026-07-07T02:19:29Z","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":"baa9097be5e017aa2b4dfb923e44ccb3cd246a986c662d9f7069a84a942393d9","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-07-06T03:51:24Z","title_canon_sha256":"1e05dd9d6850bd2fefc6dbd6c9e658b71fbe964f25be8e557452f8037cb41906"},"schema_version":"1.0","source":{"id":"2607.04640","kind":"arxiv","version":1}},"canonical_sha256":"405fa2d3ef6420b5e33ddc1639d95dffb7240f6db0efd5c1efa8de177ead74e3","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"405fa2d3ef6420b5e33ddc1639d95dffb7240f6db0efd5c1efa8de177ead74e3","first_computed_at":"2026-07-07T02:19:29.045572Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-07T02:19:29.045572Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"z9nuUrdCIUAOFgSxfe+TlLPQS/dngKSwE29L0nUn8D/aTEy9iacunIALR2NcbtW4/qooYNEkAqRdDxvUnZnmAA==","signature_status":"signed_v1","signed_at":"2026-07-07T02:19:29.046246Z","signed_message":"canonical_sha256_bytes"},"source_id":"2607.04640","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:8ad1004e6ebe577d0e2f678cdc92ca413b9c21352b047810d4743d0a72942552","sha256:7f23484653fc44b1899997397d2fa7a46e81f4f1f53ff9cae1f96fa9fe23e755"],"state_sha256":"359811d9de627ebdb4e0f8a70e36329fe0f78c7f1a1e02dfa641faa79454c8c5"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"6PSZhkrMKffVvBUmzt0sX5OEF8LaBdZ0UpCqI/x0ps5U39LqNAEJI23hQY08pB1GJwUPEUv+ntZ0uTRi9UDgBw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-15T12:39:02.304757Z","bundle_sha256":"0f78703a8cf7e0e3c5d45312bfca653c1cc4309d22f57e99335b1e3827410039"}}