{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:T2FYWMG7464GG3XM6WR3IWGPT3","short_pith_number":"pith:T2FYWMG7","schema_version":"1.0","canonical_sha256":"9e8b8b30dfe7b8636eecf5a3b458cf9ef5acdfbbb00a3ad99d653f7c58525118","source":{"kind":"arxiv","id":"2405.15624","version":2},"attestation_state":"computed","paper":{"title":"Inverse-RLignment: Large Language Model Alignment from Demonstrations through Inverse Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Hao Sun, Mihaela van der Schaar","submitted_at":"2024-05-24T15:13:53Z","abstract_excerpt":"Aligning Large Language Models (LLMs) is crucial for enhancing their safety and utility. However, existing methods, primarily based on preference datasets, face challenges such as noisy labels, high annotation costs, and privacy concerns. In this work, we introduce Alignment from Demonstrations (AfD), a novel approach leveraging high-quality demonstration data to overcome these challenges. We formalize AfD within a sequential decision-making framework, highlighting its unique challenge of missing reward signals. Drawing insights from forward and inverse reinforcement learning, we introduce div"},"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":"2405.15624","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-05-24T15:13:53Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"70c2ecb71638d8a21112d7a28845fd232b7acceb41679d8fee8a56288da5a9cb","abstract_canon_sha256":"a2e7e62b3741f41560fe8721ab2ddf69c70ce6baeb682a1078a0fa92aec30d59"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:05:10.748921Z","signature_b64":"q0PVv3bsWXxE1Ibbv5W7F0RLOZa+q/MMvKhh+hY2k1a8KFzXZIS6flA2A/sJkDDzlL/srPlq6fMG1lNlRh8MAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9e8b8b30dfe7b8636eecf5a3b458cf9ef5acdfbbb00a3ad99d653f7c58525118","last_reissued_at":"2026-07-05T10:05:10.748504Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:05:10.748504Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Inverse-RLignment: Large Language Model Alignment from Demonstrations through Inverse Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Hao Sun, Mihaela van der Schaar","submitted_at":"2024-05-24T15:13:53Z","abstract_excerpt":"Aligning Large Language Models (LLMs) is crucial for enhancing their safety and utility. However, existing methods, primarily based on preference datasets, face challenges such as noisy labels, high annotation costs, and privacy concerns. In this work, we introduce Alignment from Demonstrations (AfD), a novel approach leveraging high-quality demonstration data to overcome these challenges. We formalize AfD within a sequential decision-making framework, highlighting its unique challenge of missing reward signals. Drawing insights from forward and inverse reinforcement learning, we introduce div"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.15624","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/2405.15624/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":"2405.15624","created_at":"2026-07-05T10:05:10.748559+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.15624v2","created_at":"2026-07-05T10:05:10.748559+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.15624","created_at":"2026-07-05T10:05:10.748559+00:00"},{"alias_kind":"pith_short_12","alias_value":"T2FYWMG7464G","created_at":"2026-07-05T10:05:10.748559+00:00"},{"alias_kind":"pith_short_16","alias_value":"T2FYWMG7464GG3XM","created_at":"2026-07-05T10:05:10.748559+00:00"},{"alias_kind":"pith_short_8","alias_value":"T2FYWMG7","created_at":"2026-07-05T10:05:10.748559+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.05523","citing_title":"CHASE: Adversarial Red-Blue Teaming for Improving LLM Safety using Reinforcement Learning","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2602.07832","citing_title":"rePIRL: Learn PRM with Inverse RL for LLM Reasoning","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05710","citing_title":"On the Blessing of Pre-training in Weak-to-Strong Generalization","ref_index":107,"is_internal_anchor":false},{"citing_arxiv_id":"2604.16995","citing_title":"SPS: Steering Probability Squeezing for Better Exploration in Reinforcement Learning for Large Language Models","ref_index":31,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/T2FYWMG7464GG3XM6WR3IWGPT3","json":"https://pith.science/pith/T2FYWMG7464GG3XM6WR3IWGPT3.json","graph_json":"https://pith.science/api/pith-number/T2FYWMG7464GG3XM6WR3IWGPT3/graph.json","events_json":"https://pith.science/api/pith-number/T2FYWMG7464GG3XM6WR3IWGPT3/events.json","paper":"https://pith.science/paper/T2FYWMG7"},"agent_actions":{"view_html":"https://pith.science/pith/T2FYWMG7464GG3XM6WR3IWGPT3","download_json":"https://pith.science/pith/T2FYWMG7464GG3XM6WR3IWGPT3.json","view_paper":"https://pith.science/paper/T2FYWMG7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.15624&json=true","fetch_graph":"https://pith.science/api/pith-number/T2FYWMG7464GG3XM6WR3IWGPT3/graph.json","fetch_events":"https://pith.science/api/pith-number/T2FYWMG7464GG3XM6WR3IWGPT3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/T2FYWMG7464GG3XM6WR3IWGPT3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/T2FYWMG7464GG3XM6WR3IWGPT3/action/storage_attestation","attest_author":"https://pith.science/pith/T2FYWMG7464GG3XM6WR3IWGPT3/action/author_attestation","sign_citation":"https://pith.science/pith/T2FYWMG7464GG3XM6WR3IWGPT3/action/citation_signature","submit_replication":"https://pith.science/pith/T2FYWMG7464GG3XM6WR3IWGPT3/action/replication_record"}},"created_at":"2026-07-05T10:05:10.748559+00:00","updated_at":"2026-07-05T10:05:10.748559+00:00"}