{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:SQ4IF5TY3BRGKD6F5BMVSVN2P4","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":"75c3b87e8a501f274f185f6199870b7e203119e421ae1f2b71846f8e7dc76e19","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-03-04T23:36:55Z","title_canon_sha256":"09d72fb9a9527f525d87a573e62294c876568a39a6fa759e384c3ba68e37c252"},"schema_version":"1.0","source":{"id":"2603.04678","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2603.04678","created_at":"2026-05-29T02:05:42Z"},{"alias_kind":"arxiv_version","alias_value":"2603.04678v3","created_at":"2026-05-29T02:05:42Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.04678","created_at":"2026-05-29T02:05:42Z"},{"alias_kind":"pith_short_12","alias_value":"SQ4IF5TY3BRG","created_at":"2026-05-29T02:05:42Z"},{"alias_kind":"pith_short_16","alias_value":"SQ4IF5TY3BRGKD6F","created_at":"2026-05-29T02:05:42Z"},{"alias_kind":"pith_short_8","alias_value":"SQ4IF5TY","created_at":"2026-05-29T02:05:42Z"}],"graph_snapshots":[{"event_id":"sha256:a2f35c0fc180c745857d143d46f1eb0c7ce4a280ff5a85003d30636a9dc69165","target":"graph","created_at":"2026-05-29T02:05:42Z","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":"DCO significantly improves crosslingual consistency across diverse LLMs and outperforms existing methods when training with samples of multiple languages, while complementing DPO when gold labels are available."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That a structured reward function derived directly from the LLM itself can reliably enforce true crosslingual knowledge consistency without introducing new inconsistencies or biases during optimization."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"DCO optimizes LLMs for crosslingual knowledge consistency via a structured reward derived directly from the model itself, outperforming prior methods in experiments."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Direct Consistency Optimization derives a reward from the LLM itself to produce consistent answers to the same question across languages."}],"snapshot_sha256":"23b41dd4fd1d4743d6821102c7648dea214a93948081ab4b459126d9f4079a7f"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2603.04678/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Language models often respond inconsistently to translation-equivalent prompts across languages, undermining the reliability of multilingual systems. To quantify this, we give an information-theoretic definition of crosslingual consistency as a divergence bound between a model's response distribution and its round-trip pushforward across languages. We then introduce penalized consistency optimization (PCO), a post-training procedure that couples this divergence with a Kullback-Leibler penalty to a fixed reference language model. Because direct optimization of PCO requires expensive on-policy r","authors_text":"Arianna Bisazza, Jirui Qi, Mrinmaya Sachan, Raquel Fern\\'andez, Ryan Cotterell, Tianyu Liu","cross_cats":["cs.AI"],"headline":"Direct Consistency Optimization derives a reward from the LLM itself to produce consistent answers to the same question across languages.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-03-04T23:36:55Z","title":"Post-Training Language Models for Crosslingual Consistency"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.04678","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-15T15:56:49.533431Z","id":"61f392d0-127c-43ad-b170-703d36fdd0a3","model_set":{"reader":"grok-4.3"},"one_line_summary":"DCO optimizes LLMs for crosslingual knowledge consistency via a structured reward derived directly from the model itself, outperforming prior methods in experiments.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Direct Consistency Optimization derives a reward from the LLM itself to produce consistent answers to the same question across languages.","strongest_claim":"DCO significantly improves crosslingual consistency across diverse LLMs and outperforms existing methods when training with samples of multiple languages, while complementing DPO when gold labels are available.","weakest_assumption":"That a structured reward function derived directly from the LLM itself can reliably enforce true crosslingual knowledge consistency without introducing new inconsistencies or biases during optimization."}},"verdict_id":"61f392d0-127c-43ad-b170-703d36fdd0a3"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:2a7b3b8c812d1cbd45f8812c3503db990b0afbea727d9792425eebb3287b2230","target":"record","created_at":"2026-05-29T02:05:42Z","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":"75c3b87e8a501f274f185f6199870b7e203119e421ae1f2b71846f8e7dc76e19","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-03-04T23:36:55Z","title_canon_sha256":"09d72fb9a9527f525d87a573e62294c876568a39a6fa759e384c3ba68e37c252"},"schema_version":"1.0","source":{"id":"2603.04678","kind":"arxiv","version":3}},"canonical_sha256":"943882f678d862650fc5e8595955ba7f3ca47bbaa704f4b6609af2855da68ed2","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"943882f678d862650fc5e8595955ba7f3ca47bbaa704f4b6609af2855da68ed2","first_computed_at":"2026-05-29T02:05:42.407885Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-29T02:05:42.407885Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"keGp/SVUMJ8uItKKIiRbk3kd3dWsVsh6nlitV15oCMlyO2N0bVtb4muTCV7wslqdfQ5QHHjCRS7/T5gDZuXcAQ==","signature_status":"signed_v1","signed_at":"2026-05-29T02:05:42.408602Z","signed_message":"canonical_sha256_bytes"},"source_id":"2603.04678","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:2a7b3b8c812d1cbd45f8812c3503db990b0afbea727d9792425eebb3287b2230","sha256:a2f35c0fc180c745857d143d46f1eb0c7ce4a280ff5a85003d30636a9dc69165"],"state_sha256":"aa6b94786c60dde40862cd5bf71b6cb19a68f412ea8356a93902f8d9fc1fca90"}