{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:DVGH2NGODPWQBBGGSPYUZZENXB","short_pith_number":"pith:DVGH2NGO","canonical_record":{"source":{"id":"2501.12895","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-22T14:15:46Z","cross_cats_sorted":[],"title_canon_sha256":"8d8e4cb68cd5ddc51c076effb49b2c886747ec85d1ef3a7e27221d071af16e4f","abstract_canon_sha256":"b5b8effdd67beb3ccac934b25eadd6d94b85a9aa000d486da8e081231eab1ce5"},"schema_version":"1.0"},"canonical_sha256":"1d4c7d34ce1bed0084c693f14ce48db86abe9f788842e81d6de34fb7fff2e806","source":{"kind":"arxiv","id":"2501.12895","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2501.12895","created_at":"2026-07-05T10:04:00Z"},{"alias_kind":"arxiv_version","alias_value":"2501.12895v1","created_at":"2026-07-05T10:04:00Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.12895","created_at":"2026-07-05T10:04:00Z"},{"alias_kind":"pith_short_12","alias_value":"DVGH2NGODPWQ","created_at":"2026-07-05T10:04:00Z"},{"alias_kind":"pith_short_16","alias_value":"DVGH2NGODPWQBBGG","created_at":"2026-07-05T10:04:00Z"},{"alias_kind":"pith_short_8","alias_value":"DVGH2NGO","created_at":"2026-07-05T10:04:00Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:DVGH2NGODPWQBBGGSPYUZZENXB","target":"record","payload":{"canonical_record":{"source":{"id":"2501.12895","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-22T14:15:46Z","cross_cats_sorted":[],"title_canon_sha256":"8d8e4cb68cd5ddc51c076effb49b2c886747ec85d1ef3a7e27221d071af16e4f","abstract_canon_sha256":"b5b8effdd67beb3ccac934b25eadd6d94b85a9aa000d486da8e081231eab1ce5"},"schema_version":"1.0"},"canonical_sha256":"1d4c7d34ce1bed0084c693f14ce48db86abe9f788842e81d6de34fb7fff2e806","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:04:00.217161Z","signature_b64":"0otlZVlzDjokSBhMLXTSlHboYBsBpE3HRzLtVGEtb2C9rjoVyc00Kdskc7o1lrrnmYbkwPUDCopEjyy96xfMAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1d4c7d34ce1bed0084c693f14ce48db86abe9f788842e81d6de34fb7fff2e806","last_reissued_at":"2026-07-05T10:04:00.216689Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:04:00.216689Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2501.12895","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:04:00Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"y/cO7+WJp6ntLGWLo1Wl/qdgh7WgJfs7ym0BafDCEhupQ9LfiyOq0IIdyhfzvEnpjvxeFa1/3VM1CspGDPL9AQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-09T16:58:18.081851Z"},"content_sha256":"9bd99b40e356ac41df3ce10336ae33be00a83f32390723057be3666c9fb6a789","schema_version":"1.0","event_id":"sha256:9bd99b40e356ac41df3ce10336ae33be00a83f32390723057be3666c9fb6a789"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:DVGH2NGODPWQBBGGSPYUZZENXB","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Test-Time Preference Optimization: On-the-Fly Alignment via Iterative Textual Feedback","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Linjie Li, Xiaoye Qu, Xuyang Hu, Yafu Li, Yu Cheng","submitted_at":"2025-01-22T14:15:46Z","abstract_excerpt":"Large language models (LLMs) demonstrate impressive performance but lack the flexibility to adapt to human preferences quickly without retraining. In this work, we introduce Test-time Preference Optimization (TPO), a framework that aligns LLM outputs with human preferences during inference, removing the need to update model parameters. Rather than relying on purely numerical rewards, TPO translates reward signals into textual critiques and uses them as textual rewards to iteratively refine its response. Evaluations on benchmarks covering instruction following, preference alignment, safety, and"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.12895","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/2501.12895/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":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:04:00Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Aj5qp8JdBfdrFjCzfD2qs5RyWa5Vi60rjI5HDISZUY3lnF/eUf86xy4HtHT9wKtbMxT9r7lEczkcN9Z1Dvg0CA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-09T16:58:18.082632Z"},"content_sha256":"eda7f826246a9a168abaef28fad591737f999eab7f31dfb149913ff05b978660","schema_version":"1.0","event_id":"sha256:eda7f826246a9a168abaef28fad591737f999eab7f31dfb149913ff05b978660"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/DVGH2NGODPWQBBGGSPYUZZENXB/bundle.json","state_url":"https://pith.science/pith/DVGH2NGODPWQBBGGSPYUZZENXB/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/DVGH2NGODPWQBBGGSPYUZZENXB/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":"2026-08-09T16:58:18Z","links":{"resolver":"https://pith.science/pith/DVGH2NGODPWQBBGGSPYUZZENXB","bundle":"https://pith.science/pith/DVGH2NGODPWQBBGGSPYUZZENXB/bundle.json","state":"https://pith.science/pith/DVGH2NGODPWQBBGGSPYUZZENXB/state.json","well_known_bundle":"https://pith.science/.well-known/pith/DVGH2NGODPWQBBGGSPYUZZENXB/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:DVGH2NGODPWQBBGGSPYUZZENXB","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":"b5b8effdd67beb3ccac934b25eadd6d94b85a9aa000d486da8e081231eab1ce5","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-22T14:15:46Z","title_canon_sha256":"8d8e4cb68cd5ddc51c076effb49b2c886747ec85d1ef3a7e27221d071af16e4f"},"schema_version":"1.0","source":{"id":"2501.12895","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2501.12895","created_at":"2026-07-05T10:04:00Z"},{"alias_kind":"arxiv_version","alias_value":"2501.12895v1","created_at":"2026-07-05T10:04:00Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.12895","created_at":"2026-07-05T10:04:00Z"},{"alias_kind":"pith_short_12","alias_value":"DVGH2NGODPWQ","created_at":"2026-07-05T10:04:00Z"},{"alias_kind":"pith_short_16","alias_value":"DVGH2NGODPWQBBGG","created_at":"2026-07-05T10:04:00Z"},{"alias_kind":"pith_short_8","alias_value":"DVGH2NGO","created_at":"2026-07-05T10:04:00Z"}],"graph_snapshots":[{"event_id":"sha256:eda7f826246a9a168abaef28fad591737f999eab7f31dfb149913ff05b978660","target":"graph","created_at":"2026-07-05T10:04:00Z","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":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2501.12895/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Large language models (LLMs) demonstrate impressive performance but lack the flexibility to adapt to human preferences quickly without retraining. In this work, we introduce Test-time Preference Optimization (TPO), a framework that aligns LLM outputs with human preferences during inference, removing the need to update model parameters. Rather than relying on purely numerical rewards, TPO translates reward signals into textual critiques and uses them as textual rewards to iteratively refine its response. Evaluations on benchmarks covering instruction following, preference alignment, safety, and","authors_text":"Linjie Li, Xiaoye Qu, Xuyang Hu, Yafu Li, Yu Cheng","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-22T14:15:46Z","title":"Test-Time Preference Optimization: On-the-Fly Alignment via Iterative Textual Feedback"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.12895","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:9bd99b40e356ac41df3ce10336ae33be00a83f32390723057be3666c9fb6a789","target":"record","created_at":"2026-07-05T10:04:00Z","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":"b5b8effdd67beb3ccac934b25eadd6d94b85a9aa000d486da8e081231eab1ce5","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-22T14:15:46Z","title_canon_sha256":"8d8e4cb68cd5ddc51c076effb49b2c886747ec85d1ef3a7e27221d071af16e4f"},"schema_version":"1.0","source":{"id":"2501.12895","kind":"arxiv","version":1}},"canonical_sha256":"1d4c7d34ce1bed0084c693f14ce48db86abe9f788842e81d6de34fb7fff2e806","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"1d4c7d34ce1bed0084c693f14ce48db86abe9f788842e81d6de34fb7fff2e806","first_computed_at":"2026-07-05T10:04:00.216689Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:04:00.216689Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"0otlZVlzDjokSBhMLXTSlHboYBsBpE3HRzLtVGEtb2C9rjoVyc00Kdskc7o1lrrnmYbkwPUDCopEjyy96xfMAw==","signature_status":"signed_v1","signed_at":"2026-07-05T10:04:00.217161Z","signed_message":"canonical_sha256_bytes"},"source_id":"2501.12895","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:9bd99b40e356ac41df3ce10336ae33be00a83f32390723057be3666c9fb6a789","sha256:eda7f826246a9a168abaef28fad591737f999eab7f31dfb149913ff05b978660"],"state_sha256":"6dfd34e00f77ae040c3888b63009542dd050185c3566d673dd848935f7744768"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"SH9j3S/O5+zgnVMYBp15BA7AQYjaT1P6JDBvOy4w/jfaxXusVyuHZmpGJnKMI6PASveDk976VGVv4drSH5CxDg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-09T16:58:18.089174Z","bundle_sha256":"ec5a00d4844b361c81f06d1b147e3b8f96a1d63ad13caccb63e6f7acf6b6b15d"}}