{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:Q4O5LULEOMMCUSYUHYXYHNBUY6","short_pith_number":"pith:Q4O5LULE","canonical_record":{"source":{"id":"2604.10567","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-12T10:26:41Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"3126823eead0555446e3cf9cb37fef1b3fcbefe7600e35a8e924695cd8f1619b","abstract_canon_sha256":"4eb935cda19702d74549a85de843af8a867fe59d846e7091674c1394005bbbb7"},"schema_version":"1.0"},"canonical_sha256":"871dd5d16473182a4b143e2f83b434c7b4e3f6a1b433ab029c79fc2fcd7ec660","source":{"kind":"arxiv","id":"2604.10567","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.10567","created_at":"2026-05-28T01:04:39Z"},{"alias_kind":"arxiv_version","alias_value":"2604.10567v2","created_at":"2026-05-28T01:04:39Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.10567","created_at":"2026-05-28T01:04:39Z"},{"alias_kind":"pith_short_12","alias_value":"Q4O5LULEOMMC","created_at":"2026-05-28T01:04:39Z"},{"alias_kind":"pith_short_16","alias_value":"Q4O5LULEOMMCUSYU","created_at":"2026-05-28T01:04:39Z"},{"alias_kind":"pith_short_8","alias_value":"Q4O5LULE","created_at":"2026-05-28T01:04:39Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:Q4O5LULEOMMCUSYUHYXYHNBUY6","target":"record","payload":{"canonical_record":{"source":{"id":"2604.10567","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-12T10:26:41Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"3126823eead0555446e3cf9cb37fef1b3fcbefe7600e35a8e924695cd8f1619b","abstract_canon_sha256":"4eb935cda19702d74549a85de843af8a867fe59d846e7091674c1394005bbbb7"},"schema_version":"1.0"},"canonical_sha256":"871dd5d16473182a4b143e2f83b434c7b4e3f6a1b433ab029c79fc2fcd7ec660","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T01:04:39.941511Z","signature_b64":"zvqkh49pcCzojU4wMM3aE/sGVrmiAScN3iZ/B+E97Ug3KHQpNJLKtlMW2X5SWY2+uZ9qWhdRFN6BZg8g6lHNDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"871dd5d16473182a4b143e2f83b434c7b4e3f6a1b433ab029c79fc2fcd7ec660","last_reissued_at":"2026-05-28T01:04:39.941034Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T01:04:39.941034Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2604.10567","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":"2026-05-28T01:04:39Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"iQJoKqQ/flstMOg2v0GZFqKtbw5Qytewet4Tr4lhRTwe4E7a2v6N2YmUo7+MAlMdInHNE8IRRBJOax6Y8i1bAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-29T16:34:24.185051Z"},"content_sha256":"ee92781c02bc603267362145518c8baef808eaa2451500b8547e46cdb514fc38","schema_version":"1.0","event_id":"sha256:ee92781c02bc603267362145518c8baef808eaa2451500b8547e46cdb514fc38"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:Q4O5LULEOMMCUSYUHYXYHNBUY6","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Early Decisions Matter: Proximity Bias and Initial Trajectory Shaping in Non-Autoregressive Diffusion Language Models","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"Proximity bias in non-autoregressive diffusion language models makes the full generation trajectory depend on the position of the first unmasked token.","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Jiyeon Kim, Minjoon Seo, Moontae Lee, Sungik Choi, Yongrae Jo","submitted_at":"2026-04-12T10:26:41Z","abstract_excerpt":"Diffusion-based language models (dLLMs) have emerged as a promising alternative to autoregressive language models, offering the potential for parallel token generation and bidirectional context modeling. However, harnessing this flexibility for fully non-autoregressive decoding remains an open question, particularly for reasoning and planning tasks. In this work, we investigate non-autoregressive decoding in dLLMs by systematically analyzing its inference dynamics along the temporal axis. Specifically, we uncover an inherent failure mode in confidence-based non-autoregressive generation stemmi"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"we uncover an inherent failure mode in confidence-based non-autoregressive generation stemming from a strong proximity bias-the tendency for the denoising order to concentrate on spatially adjacent tokens. This local dependency leads to spatial error propagation, rendering the entire trajectory critically contingent on the initial unmasking position.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the observed proximity bias is the dominant cause of poor non-autoregressive performance and that a lightweight planner plus temperature annealing will reliably correct it across tasks without introducing new failure modes or significant overhead.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Non-autoregressive diffusion language models have an inherent proximity bias in token unmasking that causes spatial error propagation, which a minimal planner and annealing strategy can mitigate for better reasoning performance.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Proximity bias in non-autoregressive diffusion language models makes the full generation trajectory depend on the position of the first unmasked token.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"4984128a4abdb318090bbdc61a98cb8ff315fa565b7d5619629bed1bc133186d"},"source":{"id":"2604.10567","kind":"arxiv","version":2},"verdict":{"id":"cb3c2454-38ce-4907-b84e-142f51b76aa1","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T15:10:04.320496Z","strongest_claim":"we uncover an inherent failure mode in confidence-based non-autoregressive generation stemming from a strong proximity bias-the tendency for the denoising order to concentrate on spatially adjacent tokens. This local dependency leads to spatial error propagation, rendering the entire trajectory critically contingent on the initial unmasking position.","one_line_summary":"Non-autoregressive diffusion language models have an inherent proximity bias in token unmasking that causes spatial error propagation, which a minimal planner and annealing strategy can mitigate for better reasoning performance.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the observed proximity bias is the dominant cause of poor non-autoregressive performance and that a lightweight planner plus temperature annealing will reliably correct it across tasks without introducing new failure modes or significant overhead.","pith_extraction_headline":"Proximity bias in non-autoregressive diffusion language models makes the full generation trajectory depend on the position of the first unmasked token."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.10567/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":"cb3c2454-38ce-4907-b84e-142f51b76aa1"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-28T01:04:39Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"tit/ZLc0mWF1t4dRRmhHQYi+mdutgwo8X+muolP39P91UCQ9ywDMInS21u25Pzje13wHqCh4K8wj4gD1PZdbDw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-29T16:34:24.185519Z"},"content_sha256":"497904139cae95a0253dd9af0156b8e379c63b6ea35d719a5a3ca88190966f15","schema_version":"1.0","event_id":"sha256:497904139cae95a0253dd9af0156b8e379c63b6ea35d719a5a3ca88190966f15"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/Q4O5LULEOMMCUSYUHYXYHNBUY6/bundle.json","state_url":"https://pith.science/pith/Q4O5LULEOMMCUSYUHYXYHNBUY6/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/Q4O5LULEOMMCUSYUHYXYHNBUY6/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":"2026-06-29T16:34:24Z","links":{"resolver":"https://pith.science/pith/Q4O5LULEOMMCUSYUHYXYHNBUY6","bundle":"https://pith.science/pith/Q4O5LULEOMMCUSYUHYXYHNBUY6/bundle.json","state":"https://pith.science/pith/Q4O5LULEOMMCUSYUHYXYHNBUY6/state.json","well_known_bundle":"https://pith.science/.well-known/pith/Q4O5LULEOMMCUSYUHYXYHNBUY6/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:Q4O5LULEOMMCUSYUHYXYHNBUY6","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":"4eb935cda19702d74549a85de843af8a867fe59d846e7091674c1394005bbbb7","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-12T10:26:41Z","title_canon_sha256":"3126823eead0555446e3cf9cb37fef1b3fcbefe7600e35a8e924695cd8f1619b"},"schema_version":"1.0","source":{"id":"2604.10567","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.10567","created_at":"2026-05-28T01:04:39Z"},{"alias_kind":"arxiv_version","alias_value":"2604.10567v2","created_at":"2026-05-28T01:04:39Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.10567","created_at":"2026-05-28T01:04:39Z"},{"alias_kind":"pith_short_12","alias_value":"Q4O5LULEOMMC","created_at":"2026-05-28T01:04:39Z"},{"alias_kind":"pith_short_16","alias_value":"Q4O5LULEOMMCUSYU","created_at":"2026-05-28T01:04:39Z"},{"alias_kind":"pith_short_8","alias_value":"Q4O5LULE","created_at":"2026-05-28T01:04:39Z"}],"graph_snapshots":[{"event_id":"sha256:497904139cae95a0253dd9af0156b8e379c63b6ea35d719a5a3ca88190966f15","target":"graph","created_at":"2026-05-28T01:04:39Z","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":"we uncover an inherent failure mode in confidence-based non-autoregressive generation stemming from a strong proximity bias-the tendency for the denoising order to concentrate on spatially adjacent tokens. This local dependency leads to spatial error propagation, rendering the entire trajectory critically contingent on the initial unmasking position."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the observed proximity bias is the dominant cause of poor non-autoregressive performance and that a lightweight planner plus temperature annealing will reliably correct it across tasks without introducing new failure modes or significant overhead."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Non-autoregressive diffusion language models have an inherent proximity bias in token unmasking that causes spatial error propagation, which a minimal planner and annealing strategy can mitigate for better reasoning performance."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Proximity bias in non-autoregressive diffusion language models makes the full generation trajectory depend on the position of the first unmasked token."}],"snapshot_sha256":"4984128a4abdb318090bbdc61a98cb8ff315fa565b7d5619629bed1bc133186d"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2604.10567/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Diffusion-based language models (dLLMs) have emerged as a promising alternative to autoregressive language models, offering the potential for parallel token generation and bidirectional context modeling. However, harnessing this flexibility for fully non-autoregressive decoding remains an open question, particularly for reasoning and planning tasks. In this work, we investigate non-autoregressive decoding in dLLMs by systematically analyzing its inference dynamics along the temporal axis. Specifically, we uncover an inherent failure mode in confidence-based non-autoregressive generation stemmi","authors_text":"Jiyeon Kim, Minjoon Seo, Moontae Lee, Sungik Choi, Yongrae Jo","cross_cats":["cs.AI"],"headline":"Proximity bias in non-autoregressive diffusion language models makes the full generation trajectory depend on the position of the first unmasked token.","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-12T10:26:41Z","title":"Early Decisions Matter: Proximity Bias and Initial Trajectory Shaping in Non-Autoregressive Diffusion Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.10567","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-10T15:10:04.320496Z","id":"cb3c2454-38ce-4907-b84e-142f51b76aa1","model_set":{"reader":"grok-4.3"},"one_line_summary":"Non-autoregressive diffusion language models have an inherent proximity bias in token unmasking that causes spatial error propagation, which a minimal planner and annealing strategy can mitigate for better reasoning performance.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Proximity bias in non-autoregressive diffusion language models makes the full generation trajectory depend on the position of the first unmasked token.","strongest_claim":"we uncover an inherent failure mode in confidence-based non-autoregressive generation stemming from a strong proximity bias-the tendency for the denoising order to concentrate on spatially adjacent tokens. This local dependency leads to spatial error propagation, rendering the entire trajectory critically contingent on the initial unmasking position.","weakest_assumption":"That the observed proximity bias is the dominant cause of poor non-autoregressive performance and that a lightweight planner plus temperature annealing will reliably correct it across tasks without introducing new failure modes or significant overhead."}},"verdict_id":"cb3c2454-38ce-4907-b84e-142f51b76aa1"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:ee92781c02bc603267362145518c8baef808eaa2451500b8547e46cdb514fc38","target":"record","created_at":"2026-05-28T01:04:39Z","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":"4eb935cda19702d74549a85de843af8a867fe59d846e7091674c1394005bbbb7","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-12T10:26:41Z","title_canon_sha256":"3126823eead0555446e3cf9cb37fef1b3fcbefe7600e35a8e924695cd8f1619b"},"schema_version":"1.0","source":{"id":"2604.10567","kind":"arxiv","version":2}},"canonical_sha256":"871dd5d16473182a4b143e2f83b434c7b4e3f6a1b433ab029c79fc2fcd7ec660","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"871dd5d16473182a4b143e2f83b434c7b4e3f6a1b433ab029c79fc2fcd7ec660","first_computed_at":"2026-05-28T01:04:39.941034Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-28T01:04:39.941034Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"zvqkh49pcCzojU4wMM3aE/sGVrmiAScN3iZ/B+E97Ug3KHQpNJLKtlMW2X5SWY2+uZ9qWhdRFN6BZg8g6lHNDA==","signature_status":"signed_v1","signed_at":"2026-05-28T01:04:39.941511Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.10567","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:ee92781c02bc603267362145518c8baef808eaa2451500b8547e46cdb514fc38","sha256:497904139cae95a0253dd9af0156b8e379c63b6ea35d719a5a3ca88190966f15"],"state_sha256":"ea6802a42eab0ccb3ee475da98e017f029427414733a10ff2148c9890212e4c7"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"RkYTDdJFPHqy7EsCuLhGlDbzI18vLLxe6HGdDnCymcJnW4EnwCVYa3zPHsFKDdiB+TCcCzv36C9VYrYlFYRKCA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-29T16:34:24.187749Z","bundle_sha256":"7f057fcab7f20059e2f7b9faf1ccd81212d6d2d186c52e133e72462789f627bd"}}