{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:5DSXIWYO6NDEDXU7GWLIP3KMN4","short_pith_number":"pith:5DSXIWYO","canonical_record":{"source":{"id":"2604.08564","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-03-18T07:49:13Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"52a3bbb70a5797a703be1ddcb8f10f43fe2e08bf3191798e796d3149e5eb2d9e","abstract_canon_sha256":"34bde86f3de4a41ec2b82baeffa1d6b46adbc08c87c0c1b6b7ce839d7b8aaf6d"},"schema_version":"1.0"},"canonical_sha256":"e8e5745b0ef34641de9f359687ed4c6f2a0bf00c37e1ae46d8d057ed29df4b92","source":{"kind":"arxiv","id":"2604.08564","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.08564","created_at":"2026-06-04T01:08:49Z"},{"alias_kind":"arxiv_version","alias_value":"2604.08564v2","created_at":"2026-06-04T01:08:49Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.08564","created_at":"2026-06-04T01:08:49Z"},{"alias_kind":"pith_short_12","alias_value":"5DSXIWYO6NDE","created_at":"2026-06-04T01:08:49Z"},{"alias_kind":"pith_short_16","alias_value":"5DSXIWYO6NDEDXU7","created_at":"2026-06-04T01:08:49Z"},{"alias_kind":"pith_short_8","alias_value":"5DSXIWYO","created_at":"2026-06-04T01:08:49Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:5DSXIWYO6NDEDXU7GWLIP3KMN4","target":"record","payload":{"canonical_record":{"source":{"id":"2604.08564","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-03-18T07:49:13Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"52a3bbb70a5797a703be1ddcb8f10f43fe2e08bf3191798e796d3149e5eb2d9e","abstract_canon_sha256":"34bde86f3de4a41ec2b82baeffa1d6b46adbc08c87c0c1b6b7ce839d7b8aaf6d"},"schema_version":"1.0"},"canonical_sha256":"e8e5745b0ef34641de9f359687ed4c6f2a0bf00c37e1ae46d8d057ed29df4b92","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-04T01:08:49.742537Z","signature_b64":"b/JZ+Xn6akwSXasEsdkg+p4VUEnXS3PkbMoV3dpIaRMqc4coDsoJHFM8xLUYO/Z6GAKUSoD6CfdXE5SLTcvrAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e8e5745b0ef34641de9f359687ed4c6f2a0bf00c37e1ae46d8d057ed29df4b92","last_reissued_at":"2026-06-04T01:08:49.741875Z","signature_status":"signed_v1","first_computed_at":"2026-06-04T01:08:49.741875Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2604.08564","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":"2026-06-04T01:08:49Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"BEb9kdy6HHK3esKcP4eq99EmIVaJQv+oaz/Md4DAhJHeRIS0fB4CdGgzddpLaYpO7MXRqyI86vmIvZJhwWmSCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-06T03:07:24.685851Z"},"content_sha256":"d2832233411a431d99b6cdee39520adc2b48e8d9035442762480bdcd6e4d822a","schema_version":"1.0","event_id":"sha256:d2832233411a431d99b6cdee39520adc2b48e8d9035442762480bdcd6e4d822a"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:5DSXIWYO6NDEDXU7GWLIP3KMN4","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Attention-Based Sampler for Diffusion Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Decoding tokens in descending order of attention matrix column sums approximately maximizes sequence likelihood in diffusion language models.","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"James Kwok, Kai Syun Hou, Weiyu Chen, Yuyan Zhou","submitted_at":"2026-03-18T07:49:13Z","abstract_excerpt":"Auto-regressive models (ARMs) have established a dominant paradigm in language modeling. However, their strictly sequential sampling paradigm imposes fundamental constraints on both inference efficiency and modeling flexibility. To address these limitations, diffusion-based large language models (dLLMs) have been proposed, offering the potential for parallel sampling and flexible language modeling. Despite these advantages, current dLLMs sampling strategies rely primarily on token level information, which fails to account for global sequence structure and often yields suboptimal results. In th"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We theoretically demonstrate that optimal sequence likelihood can be approximately achieved by decoding tokens in descending order of their attention matrix column sums.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the attention matrix column sums computed during the diffusion process serve as a reliable proxy for each token's marginal contribution to the joint log-likelihood, without requiring additional fitting or model-specific adjustments.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Attn-Sampler decodes diffusion language models by selecting tokens in descending order of attention column sums, yielding higher quality and more parallel generation than token-level greedy baselines.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Decoding tokens in descending order of attention matrix column sums approximately maximizes sequence likelihood in diffusion language models.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"ad16f8d22f2dcd23882c82b58e9fa9ea179386157e2b1084232ce360d8829054"},"source":{"id":"2604.08564","kind":"arxiv","version":2},"verdict":{"id":"871f4d78-9a11-4202-b2ee-415d9eb7a024","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T10:13:55.782710Z","strongest_claim":"We theoretically demonstrate that optimal sequence likelihood can be approximately achieved by decoding tokens in descending order of their attention matrix column sums.","one_line_summary":"Attn-Sampler decodes diffusion language models by selecting tokens in descending order of attention column sums, yielding higher quality and more parallel generation than token-level greedy baselines.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the attention matrix column sums computed during the diffusion process serve as a reliable proxy for each token's marginal contribution to the joint log-likelihood, without requiring additional fitting or model-specific adjustments.","pith_extraction_headline":"Decoding tokens in descending order of attention matrix column sums approximately maximizes sequence likelihood in diffusion language models."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.08564/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":"871f4d78-9a11-4202-b2ee-415d9eb7a024"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-04T01:08:49Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"/lHV8aL3UXvpSLl4DKUPJDJW6o67Q3LtWs5Hz/TJoKfQXFNCYSocaVdtOxt8YObr6iY3e+rBc1UaxTbxeyo9DQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-06T03:07:24.686228Z"},"content_sha256":"86d1006791146f136d26924ce2c833f00e937a0740ac1cd082f2b6383077d318","schema_version":"1.0","event_id":"sha256:86d1006791146f136d26924ce2c833f00e937a0740ac1cd082f2b6383077d318"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/5DSXIWYO6NDEDXU7GWLIP3KMN4/bundle.json","state_url":"https://pith.science/pith/5DSXIWYO6NDEDXU7GWLIP3KMN4/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/5DSXIWYO6NDEDXU7GWLIP3KMN4/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":"2026-08-06T03:07:24Z","links":{"resolver":"https://pith.science/pith/5DSXIWYO6NDEDXU7GWLIP3KMN4","bundle":"https://pith.science/pith/5DSXIWYO6NDEDXU7GWLIP3KMN4/bundle.json","state":"https://pith.science/pith/5DSXIWYO6NDEDXU7GWLIP3KMN4/state.json","well_known_bundle":"https://pith.science/.well-known/pith/5DSXIWYO6NDEDXU7GWLIP3KMN4/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:5DSXIWYO6NDEDXU7GWLIP3KMN4","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":"34bde86f3de4a41ec2b82baeffa1d6b46adbc08c87c0c1b6b7ce839d7b8aaf6d","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-03-18T07:49:13Z","title_canon_sha256":"52a3bbb70a5797a703be1ddcb8f10f43fe2e08bf3191798e796d3149e5eb2d9e"},"schema_version":"1.0","source":{"id":"2604.08564","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.08564","created_at":"2026-06-04T01:08:49Z"},{"alias_kind":"arxiv_version","alias_value":"2604.08564v2","created_at":"2026-06-04T01:08:49Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.08564","created_at":"2026-06-04T01:08:49Z"},{"alias_kind":"pith_short_12","alias_value":"5DSXIWYO6NDE","created_at":"2026-06-04T01:08:49Z"},{"alias_kind":"pith_short_16","alias_value":"5DSXIWYO6NDEDXU7","created_at":"2026-06-04T01:08:49Z"},{"alias_kind":"pith_short_8","alias_value":"5DSXIWYO","created_at":"2026-06-04T01:08:49Z"}],"graph_snapshots":[{"event_id":"sha256:86d1006791146f136d26924ce2c833f00e937a0740ac1cd082f2b6383077d318","target":"graph","created_at":"2026-06-04T01:08:49Z","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 theoretically demonstrate that optimal sequence likelihood can be approximately achieved by decoding tokens in descending order of their attention matrix column sums."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the attention matrix column sums computed during the diffusion process serve as a reliable proxy for each token's marginal contribution to the joint log-likelihood, without requiring additional fitting or model-specific adjustments."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Attn-Sampler decodes diffusion language models by selecting tokens in descending order of attention column sums, yielding higher quality and more parallel generation than token-level greedy baselines."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Decoding tokens in descending order of attention matrix column sums approximately maximizes sequence likelihood in diffusion language models."}],"snapshot_sha256":"ad16f8d22f2dcd23882c82b58e9fa9ea179386157e2b1084232ce360d8829054"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2604.08564/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Auto-regressive models (ARMs) have established a dominant paradigm in language modeling. However, their strictly sequential sampling paradigm imposes fundamental constraints on both inference efficiency and modeling flexibility. To address these limitations, diffusion-based large language models (dLLMs) have been proposed, offering the potential for parallel sampling and flexible language modeling. Despite these advantages, current dLLMs sampling strategies rely primarily on token level information, which fails to account for global sequence structure and often yields suboptimal results. In th","authors_text":"James Kwok, Kai Syun Hou, Weiyu Chen, Yuyan Zhou","cross_cats":["cs.LG"],"headline":"Decoding tokens in descending order of attention matrix column sums approximately maximizes sequence likelihood in diffusion language models.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-03-18T07:49:13Z","title":"Attention-Based Sampler for Diffusion Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.08564","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-15T10:13:55.782710Z","id":"871f4d78-9a11-4202-b2ee-415d9eb7a024","model_set":{"reader":"grok-4.3"},"one_line_summary":"Attn-Sampler decodes diffusion language models by selecting tokens in descending order of attention column sums, yielding higher quality and more parallel generation than token-level greedy baselines.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Decoding tokens in descending order of attention matrix column sums approximately maximizes sequence likelihood in diffusion language models.","strongest_claim":"We theoretically demonstrate that optimal sequence likelihood can be approximately achieved by decoding tokens in descending order of their attention matrix column sums.","weakest_assumption":"That the attention matrix column sums computed during the diffusion process serve as a reliable proxy for each token's marginal contribution to the joint log-likelihood, without requiring additional fitting or model-specific adjustments."}},"verdict_id":"871f4d78-9a11-4202-b2ee-415d9eb7a024"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:d2832233411a431d99b6cdee39520adc2b48e8d9035442762480bdcd6e4d822a","target":"record","created_at":"2026-06-04T01:08:49Z","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":"34bde86f3de4a41ec2b82baeffa1d6b46adbc08c87c0c1b6b7ce839d7b8aaf6d","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-03-18T07:49:13Z","title_canon_sha256":"52a3bbb70a5797a703be1ddcb8f10f43fe2e08bf3191798e796d3149e5eb2d9e"},"schema_version":"1.0","source":{"id":"2604.08564","kind":"arxiv","version":2}},"canonical_sha256":"e8e5745b0ef34641de9f359687ed4c6f2a0bf00c37e1ae46d8d057ed29df4b92","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"e8e5745b0ef34641de9f359687ed4c6f2a0bf00c37e1ae46d8d057ed29df4b92","first_computed_at":"2026-06-04T01:08:49.741875Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-04T01:08:49.741875Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"b/JZ+Xn6akwSXasEsdkg+p4VUEnXS3PkbMoV3dpIaRMqc4coDsoJHFM8xLUYO/Z6GAKUSoD6CfdXE5SLTcvrAQ==","signature_status":"signed_v1","signed_at":"2026-06-04T01:08:49.742537Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.08564","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:d2832233411a431d99b6cdee39520adc2b48e8d9035442762480bdcd6e4d822a","sha256:86d1006791146f136d26924ce2c833f00e937a0740ac1cd082f2b6383077d318"],"state_sha256":"f0176a443935a0f8a52016ce7b1e22d94b46a5ffde44cefcb0c279c19c9994b0"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"e7Ldt9j04wS0ivRQ6o4x0ctTtGo4JITTjEzCuzx+YAFv/wed3oOo8ZkQKrujv17+Tuv4fMtcNjl0HBmP+GDoCg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-06T03:07:24.689478Z","bundle_sha256":"48f76b6ea9dfe089578b6467a5e48b07847529c238b593f240888a50f8393e38"}}