{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:FHXUMIEKFCKMMRYQ5VB3H26VSE","short_pith_number":"pith:FHXUMIEK","schema_version":"1.0","canonical_sha256":"29ef46208a2894c64710ed43b3ebd5913f1406d4c3ba7d4c75de612c9832ec84","source":{"kind":"arxiv","id":"2605.14465","version":1},"attestation_state":"computed","paper":{"title":"From Table to Cell: Attention for Better Reasoning with TABALIGN","license":"http://creativecommons.org/licenses/by/4.0/","headline":"TABALIGN pairs a diffusion language model planner that emits binary cell masks with an attention verifier trained on human standards to enforce cell grounding in table reasoning.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Chunhe Wang, Guang Cheng, Hanwei Wu, Lei Ding, Tung Sum Thomas Kwok, Xiaofeng Lin, Xinyu Wang, Zeyong Zhang, Zhijiang Guo","submitted_at":"2026-05-14T07:00:26Z","abstract_excerpt":"Multi-step LLM reasoning over structured tables fails because planning and execution share no explicit cell-grounding contract. Existing methods constrain the planner to a left-to-right factorization at odds with table permutation invariance, and score intermediate states by generated content alone, overlooking cell grounding. We conduct a pilot study showing that diffusion language models (DLMs) produce more human-aligned and permutation-stable cell attention on tables than autoregressive models, with a 40.2% median reduction in attention-AUROC variability under row reordering. Motivated by t"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":true,"formal_links_present":false},"canonical_record":{"source":{"id":"2605.14465","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-14T07:00:26Z","cross_cats_sorted":[],"title_canon_sha256":"d621ba3c98dd9262c7ecb24095d95a674207be560406ebfead348bb859db2e4a","abstract_canon_sha256":"9e7f88313850a5eaee3ab810d62ea492776f67bf233a5c15e27493754773f730"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:06.732158Z","signature_b64":"wn95A7Lx2xKQjVZ+/0yEctJjO53uXChKXofCBbWjdI2g3kFG5TqEins6akMw7f7HbQ1lXGp+gdeMqdqxddkaAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"29ef46208a2894c64710ed43b3ebd5913f1406d4c3ba7d4c75de612c9832ec84","last_reissued_at":"2026-05-17T23:39:06.731364Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:06.731364Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"From Table to Cell: Attention for Better Reasoning with TABALIGN","license":"http://creativecommons.org/licenses/by/4.0/","headline":"TABALIGN pairs a diffusion language model planner that emits binary cell masks with an attention verifier trained on human standards to enforce cell grounding in table reasoning.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Chunhe Wang, Guang Cheng, Hanwei Wu, Lei Ding, Tung Sum Thomas Kwok, Xiaofeng Lin, Xinyu Wang, Zeyong Zhang, Zhijiang Guo","submitted_at":"2026-05-14T07:00:26Z","abstract_excerpt":"Multi-step LLM reasoning over structured tables fails because planning and execution share no explicit cell-grounding contract. Existing methods constrain the planner to a left-to-right factorization at odds with table permutation invariance, and score intermediate states by generated content alone, overlooking cell grounding. We conduct a pilot study showing that diffusion language models (DLMs) produce more human-aligned and permutation-stable cell attention on tables than autoregressive models, with a 40.2% median reduction in attention-AUROC variability under row reordering. Motivated by t"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Across eight benchmarks covering table question answering and fact verification, TABALIGN improves average accuracy by 15.76 percentage points over the strongest open-source baseline at comparable 8B-class scale, with a matched-backbone ablation attributing 2.87 percentage points of this gain to the DLM planner over an AR planner on a fixed reasoner.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The 1,600 human-verified attention standards used to train TABATTN are representative and stable across different table layouts, domains, and model scales; if they are not, the verifier's scoring may not reliably enforce the cell-grounding contract.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"TABALIGN pairs a diffusion language model planner emitting binary cell masks with a trained attention verifier, raising average accuracy 15.76 points over strong baselines on eight table benchmarks while speeding execution 44.64%.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"TABALIGN pairs a diffusion language model planner that emits binary cell masks with an attention verifier trained on human standards to enforce cell grounding in table reasoning.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"c3522d648c2513534933f5c268480c369bc6378ce1d392ac19867ded80cb822d"},"source":{"id":"2605.14465","kind":"arxiv","version":1},"verdict":{"id":"650f87fb-fe46-407e-b4bd-8ef7b50b9cb6","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T01:25:47.957637Z","strongest_claim":"Across eight benchmarks covering table question answering and fact verification, TABALIGN improves average accuracy by 15.76 percentage points over the strongest open-source baseline at comparable 8B-class scale, with a matched-backbone ablation attributing 2.87 percentage points of this gain to the DLM planner over an AR planner on a fixed reasoner.","one_line_summary":"TABALIGN pairs a diffusion language model planner emitting binary cell masks with a trained attention verifier, raising average accuracy 15.76 points over strong baselines on eight table benchmarks while speeding execution 44.64%.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The 1,600 human-verified attention standards used to train TABATTN are representative and stable across different table layouts, domains, and model scales; if they are not, the verifier's scoring may not reliably enforce the cell-grounding contract.","pith_extraction_headline":"TABALIGN pairs a diffusion language model planner that emits binary cell masks with an attention verifier trained on human standards to enforce cell grounding in table reasoning."},"references":{"count":87,"sample":[{"doi":"","year":2025,"title":"Adapting autoregressive vision language models for parallel diffusion decoding, 2025","work_id":"7ffb24b6-5db0-46fd-a761-f5611015ec40","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Johnson, Jonathan Ho, Daniel Tarlow, and Rianne van den Berg","work_id":"368943f8-485f-4c02-aa7b-601f5fc8ce36","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"URLhttps://openreview.net/forum?id=h7-XixPCAL","work_id":"54befcc6-ed61-456b-bfa5-58d9d4a3238e","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Qwen3-VL Technical Report","work_id":"1fe243aa-e3c0-4da6-b391-4cbcfc88d5c0","ref_index":4,"cited_arxiv_id":"2511.21631","is_internal_anchor":true},{"doi":"","year":2025,"title":"Qwen2.5-VL Technical Report","work_id":"69dffacb-bfe8-442d-be86-48624c60426f","ref_index":5,"cited_arxiv_id":"2502.13923","is_internal_anchor":true}],"resolved_work":87,"snapshot_sha256":"bff895ffc51f7c669858c88e4a91ae9781cccab955420665c020172f61480284","internal_anchors":15},"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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2605.14465","created_at":"2026-05-17T23:39:06.731491+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.14465v1","created_at":"2026-05-17T23:39:06.731491+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.14465","created_at":"2026-05-17T23:39:06.731491+00:00"},{"alias_kind":"pith_short_12","alias_value":"FHXUMIEKFCKM","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"FHXUMIEKFCKMMRYQ","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"FHXUMIEK","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FHXUMIEKFCKMMRYQ5VB3H26VSE","json":"https://pith.science/pith/FHXUMIEKFCKMMRYQ5VB3H26VSE.json","graph_json":"https://pith.science/api/pith-number/FHXUMIEKFCKMMRYQ5VB3H26VSE/graph.json","events_json":"https://pith.science/api/pith-number/FHXUMIEKFCKMMRYQ5VB3H26VSE/events.json","paper":"https://pith.science/paper/FHXUMIEK"},"agent_actions":{"view_html":"https://pith.science/pith/FHXUMIEKFCKMMRYQ5VB3H26VSE","download_json":"https://pith.science/pith/FHXUMIEKFCKMMRYQ5VB3H26VSE.json","view_paper":"https://pith.science/paper/FHXUMIEK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.14465&json=true","fetch_graph":"https://pith.science/api/pith-number/FHXUMIEKFCKMMRYQ5VB3H26VSE/graph.json","fetch_events":"https://pith.science/api/pith-number/FHXUMIEKFCKMMRYQ5VB3H26VSE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FHXUMIEKFCKMMRYQ5VB3H26VSE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FHXUMIEKFCKMMRYQ5VB3H26VSE/action/storage_attestation","attest_author":"https://pith.science/pith/FHXUMIEKFCKMMRYQ5VB3H26VSE/action/author_attestation","sign_citation":"https://pith.science/pith/FHXUMIEKFCKMMRYQ5VB3H26VSE/action/citation_signature","submit_replication":"https://pith.science/pith/FHXUMIEKFCKMMRYQ5VB3H26VSE/action/replication_record"}},"created_at":"2026-05-17T23:39:06.731491+00:00","updated_at":"2026-05-17T23:39:06.731491+00:00"}