{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:CA7QTAARYZ6UD4VFVKXPRMPLFZ","short_pith_number":"pith:CA7QTAAR","schema_version":"1.0","canonical_sha256":"103f098011c67d41f2a5aaaef8b1eb2e510f7409c3b5dbdbc7e09a79b4066bbd","source":{"kind":"arxiv","id":"2502.01567","version":2},"attestation_state":"computed","paper":{"title":"Latent Thought Models with Variational Bayes Inference-Time Computation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG","stat.ML"],"primary_cat":"cs.CL","authors_text":"Bo Pang, Chuan Li, Dehong Xu, Deqian Kong, Edouardo Honig, Jianwen Xie, Minglu Zhao, Shu Wang, Sirui Xie, Ying Nian Wu, Zhangzhang Si","submitted_at":"2025-02-03T17:50:34Z","abstract_excerpt":"We propose a novel class of language models, Latent Thought Models (LTMs), which incorporate explicit latent thought vectors that follow an explicit prior model in latent space. These latent thought vectors guide the autoregressive generation of ground tokens through a Transformer decoder. Training employs a dual-rate optimization process within the classical variational Bayes framework: fast learning of local variational parameters for the posterior distribution of latent vectors (inference-time computation), and slow learning of global decoder parameters. Empirical studies reveal that LTMs p"},"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":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2502.01567","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-03T17:50:34Z","cross_cats_sorted":["cs.LG","stat.ML"],"title_canon_sha256":"70869a5e3d1bd2d3ac04d62f8c1d4236d978541df2cc56bd713323482c0f1cbd","abstract_canon_sha256":"87856941f5fde9a4b2af7c09d49325874dff94c799cf57b6890a714ea82debb6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:17:40.733397Z","signature_b64":"CIAX29zdVmJcd6UaXcETl4ve1dOmfwSeS1kb1pbFuR2uye0oqb2WNWvG7bg5fI/aMJ+2mwlVepT5e4VxrjnrCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"103f098011c67d41f2a5aaaef8b1eb2e510f7409c3b5dbdbc7e09a79b4066bbd","last_reissued_at":"2026-07-05T11:17:40.732896Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:17:40.732896Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Latent Thought Models with Variational Bayes Inference-Time Computation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG","stat.ML"],"primary_cat":"cs.CL","authors_text":"Bo Pang, Chuan Li, Dehong Xu, Deqian Kong, Edouardo Honig, Jianwen Xie, Minglu Zhao, Shu Wang, Sirui Xie, Ying Nian Wu, Zhangzhang Si","submitted_at":"2025-02-03T17:50:34Z","abstract_excerpt":"We propose a novel class of language models, Latent Thought Models (LTMs), which incorporate explicit latent thought vectors that follow an explicit prior model in latent space. These latent thought vectors guide the autoregressive generation of ground tokens through a Transformer decoder. Training employs a dual-rate optimization process within the classical variational Bayes framework: fast learning of local variational parameters for the posterior distribution of latent vectors (inference-time computation), and slow learning of global decoder parameters. Empirical studies reveal that LTMs p"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.01567","kind":"arxiv","version":2},"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/2502.01567/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2502.01567","created_at":"2026-07-05T11:17:40.732951+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.01567v2","created_at":"2026-07-05T11:17:40.732951+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.01567","created_at":"2026-07-05T11:17:40.732951+00:00"},{"alias_kind":"pith_short_12","alias_value":"CA7QTAARYZ6U","created_at":"2026-07-05T11:17:40.732951+00:00"},{"alias_kind":"pith_short_16","alias_value":"CA7QTAARYZ6UD4VF","created_at":"2026-07-05T11:17:40.732951+00:00"},{"alias_kind":"pith_short_8","alias_value":"CA7QTAAR","created_at":"2026-07-05T11:17:40.732951+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2505.17384","citing_title":"Variational Autoencoding Discrete Diffusion with Enhanced Dimensional Correlations Modeling","ref_index":7,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CA7QTAARYZ6UD4VFVKXPRMPLFZ","json":"https://pith.science/pith/CA7QTAARYZ6UD4VFVKXPRMPLFZ.json","graph_json":"https://pith.science/api/pith-number/CA7QTAARYZ6UD4VFVKXPRMPLFZ/graph.json","events_json":"https://pith.science/api/pith-number/CA7QTAARYZ6UD4VFVKXPRMPLFZ/events.json","paper":"https://pith.science/paper/CA7QTAAR"},"agent_actions":{"view_html":"https://pith.science/pith/CA7QTAARYZ6UD4VFVKXPRMPLFZ","download_json":"https://pith.science/pith/CA7QTAARYZ6UD4VFVKXPRMPLFZ.json","view_paper":"https://pith.science/paper/CA7QTAAR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.01567&json=true","fetch_graph":"https://pith.science/api/pith-number/CA7QTAARYZ6UD4VFVKXPRMPLFZ/graph.json","fetch_events":"https://pith.science/api/pith-number/CA7QTAARYZ6UD4VFVKXPRMPLFZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CA7QTAARYZ6UD4VFVKXPRMPLFZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CA7QTAARYZ6UD4VFVKXPRMPLFZ/action/storage_attestation","attest_author":"https://pith.science/pith/CA7QTAARYZ6UD4VFVKXPRMPLFZ/action/author_attestation","sign_citation":"https://pith.science/pith/CA7QTAARYZ6UD4VFVKXPRMPLFZ/action/citation_signature","submit_replication":"https://pith.science/pith/CA7QTAARYZ6UD4VFVKXPRMPLFZ/action/replication_record"}},"created_at":"2026-07-05T11:17:40.732951+00:00","updated_at":"2026-07-05T11:17:40.732951+00:00"}