{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:BTJRI6D7ONU7JDARGKQVTQNTFW","short_pith_number":"pith:BTJRI6D7","schema_version":"1.0","canonical_sha256":"0cd314787f7369f48c1132a159c1b32d9fee363ea24d3dc9207fcb32e3565b31","source":{"kind":"arxiv","id":"2404.17546","version":1},"attestation_state":"computed","paper":{"title":"Probabilistic Inference in Language Models via Twisted Sequential Monte Carlo","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","stat.ML"],"primary_cat":"cs.LG","authors_text":"Alireza Makhzani, Rob Brekelmans, Roger Grosse, Stephen Zhao","submitted_at":"2024-04-26T17:18:32Z","abstract_excerpt":"Numerous capability and safety techniques of Large Language Models (LLMs), including RLHF, automated red-teaming, prompt engineering, and infilling, can be cast as sampling from an unnormalized target distribution defined by a given reward or potential function over the full sequence. In this work, we leverage the rich toolkit of Sequential Monte Carlo (SMC) for these probabilistic inference problems. In particular, we use learned twist functions to estimate the expected future value of the potential at each timestep, which enables us to focus inference-time computation on promising partial se"},"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":"2404.17546","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-04-26T17:18:32Z","cross_cats_sorted":["cs.AI","cs.CL","stat.ML"],"title_canon_sha256":"f68580284269992fcbaea7b57994a65f538d8215c68d0f2352c1043b8125ac84","abstract_canon_sha256":"c791e1f156805016a5d962dba628b6ff7e6e38a6e0583cb3cd8245014d97a435"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:13:34.115046Z","signature_b64":"nkc5SAfXcCIYmkl4JVkTBY07PCT3RgRQxnzZw/GYji3rpHfHDRkbcmqT22zeFwOMozkavftVGGIucvKAPc+/BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0cd314787f7369f48c1132a159c1b32d9fee363ea24d3dc9207fcb32e3565b31","last_reissued_at":"2026-07-05T08:13:34.114567Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:13:34.114567Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Probabilistic Inference in Language Models via Twisted Sequential Monte Carlo","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","stat.ML"],"primary_cat":"cs.LG","authors_text":"Alireza Makhzani, Rob Brekelmans, Roger Grosse, Stephen Zhao","submitted_at":"2024-04-26T17:18:32Z","abstract_excerpt":"Numerous capability and safety techniques of Large Language Models (LLMs), including RLHF, automated red-teaming, prompt engineering, and infilling, can be cast as sampling from an unnormalized target distribution defined by a given reward or potential function over the full sequence. In this work, we leverage the rich toolkit of Sequential Monte Carlo (SMC) for these probabilistic inference problems. In particular, we use learned twist functions to estimate the expected future value of the potential at each timestep, which enables us to focus inference-time computation on promising partial se"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.17546","kind":"arxiv","version":1},"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/2404.17546/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":"2404.17546","created_at":"2026-07-05T08:13:34.114627+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.17546v1","created_at":"2026-07-05T08:13:34.114627+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.17546","created_at":"2026-07-05T08:13:34.114627+00:00"},{"alias_kind":"pith_short_12","alias_value":"BTJRI6D7ONU7","created_at":"2026-07-05T08:13:34.114627+00:00"},{"alias_kind":"pith_short_16","alias_value":"BTJRI6D7ONU7JDAR","created_at":"2026-07-05T08:13:34.114627+00:00"},{"alias_kind":"pith_short_8","alias_value":"BTJRI6D7","created_at":"2026-07-05T08:13:34.114627+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.23346","citing_title":"Contrastive Distribution Matching for Amortized Sequential Monte Carlo in Discrete Diffusion","ref_index":95,"is_internal_anchor":false},{"citing_arxiv_id":"2409.02399","citing_title":"Guidance for twisted particle filter: a continuous-time perspective","ref_index":67,"is_internal_anchor":false},{"citing_arxiv_id":"2510.12916","citing_title":"Efficient Inference for Coupled Hidden Markov Models in Continuous Time and Discrete Space","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2604.16453","citing_title":"Sampling for Quality: Training-Free Reward-Guided LLM Decoding via Sequential Monte Carlo","ref_index":20,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BTJRI6D7ONU7JDARGKQVTQNTFW","json":"https://pith.science/pith/BTJRI6D7ONU7JDARGKQVTQNTFW.json","graph_json":"https://pith.science/api/pith-number/BTJRI6D7ONU7JDARGKQVTQNTFW/graph.json","events_json":"https://pith.science/api/pith-number/BTJRI6D7ONU7JDARGKQVTQNTFW/events.json","paper":"https://pith.science/paper/BTJRI6D7"},"agent_actions":{"view_html":"https://pith.science/pith/BTJRI6D7ONU7JDARGKQVTQNTFW","download_json":"https://pith.science/pith/BTJRI6D7ONU7JDARGKQVTQNTFW.json","view_paper":"https://pith.science/paper/BTJRI6D7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.17546&json=true","fetch_graph":"https://pith.science/api/pith-number/BTJRI6D7ONU7JDARGKQVTQNTFW/graph.json","fetch_events":"https://pith.science/api/pith-number/BTJRI6D7ONU7JDARGKQVTQNTFW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BTJRI6D7ONU7JDARGKQVTQNTFW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BTJRI6D7ONU7JDARGKQVTQNTFW/action/storage_attestation","attest_author":"https://pith.science/pith/BTJRI6D7ONU7JDARGKQVTQNTFW/action/author_attestation","sign_citation":"https://pith.science/pith/BTJRI6D7ONU7JDARGKQVTQNTFW/action/citation_signature","submit_replication":"https://pith.science/pith/BTJRI6D7ONU7JDARGKQVTQNTFW/action/replication_record"}},"created_at":"2026-07-05T08:13:34.114627+00:00","updated_at":"2026-07-05T08:13:34.114627+00:00"}