{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:6DPV6LX67IXAQBSE6RYY6AKZXM","short_pith_number":"pith:6DPV6LX6","schema_version":"1.0","canonical_sha256":"f0df5f2efefa2e080644f4718f0159bb0c0076d708b381a4394ac4b9d3b21ced","source":{"kind":"arxiv","id":"2406.00024","version":2},"attestation_state":"computed","paper":{"title":"Embedding-Aligned Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.ET","cs.LG"],"primary_cat":"cs.CL","authors_text":"Chih-wei Hsu, Craig Boutilier, Ethan Liang, Guy Tennenholtz, Lior Shani, Yinlam Chow","submitted_at":"2024-05-24T06:11:17Z","abstract_excerpt":"We propose a novel approach for training large language models (LLMs) to adhere to objectives defined within a latent embedding space. Our method leverages reinforcement learning (RL), treating a pre-trained LLM as an environment. Our embedding-aligned guided language (EAGLE) agent is trained to iteratively steer the LLM's generation towards optimal regions of the latent embedding space, w.r.t. some predefined criterion. We demonstrate the effectiveness of the EAGLE agent using the MovieLens 25M and Amazon Review datasets to surface content gaps that satisfy latent user demand. We also demonst"},"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":"2406.00024","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-24T06:11:17Z","cross_cats_sorted":["cs.AI","cs.ET","cs.LG"],"title_canon_sha256":"790153c296d9f0f21fa7abab55a9363c38a5a3a24096d93513ab5af066476415","abstract_canon_sha256":"fd1deee49a1be26fab4073bfa0b3f9f3b2a1bc520b09ddd946f4e45c87966d76"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:27:01.489375Z","signature_b64":"0rPNBdbZ6Ms9ngtTQdZLO2uY23gL+yk+ZhtFYBEt02ymATV5kGglFU8O80b16bdiSQwgP5xBEuDI5ZMh5LphAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f0df5f2efefa2e080644f4718f0159bb0c0076d708b381a4394ac4b9d3b21ced","last_reissued_at":"2026-07-05T09:27:01.488960Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:27:01.488960Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Embedding-Aligned Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.ET","cs.LG"],"primary_cat":"cs.CL","authors_text":"Chih-wei Hsu, Craig Boutilier, Ethan Liang, Guy Tennenholtz, Lior Shani, Yinlam Chow","submitted_at":"2024-05-24T06:11:17Z","abstract_excerpt":"We propose a novel approach for training large language models (LLMs) to adhere to objectives defined within a latent embedding space. Our method leverages reinforcement learning (RL), treating a pre-trained LLM as an environment. Our embedding-aligned guided language (EAGLE) agent is trained to iteratively steer the LLM's generation towards optimal regions of the latent embedding space, w.r.t. some predefined criterion. We demonstrate the effectiveness of the EAGLE agent using the MovieLens 25M and Amazon Review datasets to surface content gaps that satisfy latent user demand. We also demonst"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.00024","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/2406.00024/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":"2406.00024","created_at":"2026-07-05T09:27:01.489008+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.00024v2","created_at":"2026-07-05T09:27:01.489008+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.00024","created_at":"2026-07-05T09:27:01.489008+00:00"},{"alias_kind":"pith_short_12","alias_value":"6DPV6LX67IXA","created_at":"2026-07-05T09:27:01.489008+00:00"},{"alias_kind":"pith_short_16","alias_value":"6DPV6LX67IXAQBSE","created_at":"2026-07-05T09:27:01.489008+00:00"},{"alias_kind":"pith_short_8","alias_value":"6DPV6LX6","created_at":"2026-07-05T09:27:01.489008+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2412.19396","citing_title":"Comparing Few to Rank Many: Active Human Preference Learning using Randomized Frank-Wolfe","ref_index":52,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6DPV6LX67IXAQBSE6RYY6AKZXM","json":"https://pith.science/pith/6DPV6LX67IXAQBSE6RYY6AKZXM.json","graph_json":"https://pith.science/api/pith-number/6DPV6LX67IXAQBSE6RYY6AKZXM/graph.json","events_json":"https://pith.science/api/pith-number/6DPV6LX67IXAQBSE6RYY6AKZXM/events.json","paper":"https://pith.science/paper/6DPV6LX6"},"agent_actions":{"view_html":"https://pith.science/pith/6DPV6LX67IXAQBSE6RYY6AKZXM","download_json":"https://pith.science/pith/6DPV6LX67IXAQBSE6RYY6AKZXM.json","view_paper":"https://pith.science/paper/6DPV6LX6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.00024&json=true","fetch_graph":"https://pith.science/api/pith-number/6DPV6LX67IXAQBSE6RYY6AKZXM/graph.json","fetch_events":"https://pith.science/api/pith-number/6DPV6LX67IXAQBSE6RYY6AKZXM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6DPV6LX67IXAQBSE6RYY6AKZXM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6DPV6LX67IXAQBSE6RYY6AKZXM/action/storage_attestation","attest_author":"https://pith.science/pith/6DPV6LX67IXAQBSE6RYY6AKZXM/action/author_attestation","sign_citation":"https://pith.science/pith/6DPV6LX67IXAQBSE6RYY6AKZXM/action/citation_signature","submit_replication":"https://pith.science/pith/6DPV6LX67IXAQBSE6RYY6AKZXM/action/replication_record"}},"created_at":"2026-07-05T09:27:01.489008+00:00","updated_at":"2026-07-05T09:27:01.489008+00:00"}