{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:BWRBMXALT3CSUABVG57QOVALSS","short_pith_number":"pith:BWRBMXAL","canonical_record":{"source":{"id":"2608.07419","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-08-07T17:05:10Z","cross_cats_sorted":[],"title_canon_sha256":"20d175ba28e2316001849ebbbf4d0d62ad89727c637d3ec20704fe51286be963","abstract_canon_sha256":"19a2301de88efa8b0d9f90ea54068aa63e3c88f938ffe28a36c4b0b478002bad"},"schema_version":"1.0"},"canonical_sha256":"0da2165c0b9ec52a0035377f07540b9485884d10b34c2419411a4a5d711c4b65","source":{"kind":"arxiv","id":"2608.07419","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2608.07419","created_at":"2026-08-10T01:15:20Z"},{"alias_kind":"arxiv_version","alias_value":"2608.07419v1","created_at":"2026-08-10T01:15:20Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.07419","created_at":"2026-08-10T01:15:20Z"},{"alias_kind":"pith_short_12","alias_value":"BWRBMXALT3CS","created_at":"2026-08-10T01:15:20Z"},{"alias_kind":"pith_short_16","alias_value":"BWRBMXALT3CSUABV","created_at":"2026-08-10T01:15:20Z"},{"alias_kind":"pith_short_8","alias_value":"BWRBMXAL","created_at":"2026-08-10T01:15:20Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:BWRBMXALT3CSUABVG57QOVALSS","target":"record","payload":{"canonical_record":{"source":{"id":"2608.07419","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-08-07T17:05:10Z","cross_cats_sorted":[],"title_canon_sha256":"20d175ba28e2316001849ebbbf4d0d62ad89727c637d3ec20704fe51286be963","abstract_canon_sha256":"19a2301de88efa8b0d9f90ea54068aa63e3c88f938ffe28a36c4b0b478002bad"},"schema_version":"1.0"},"canonical_sha256":"0da2165c0b9ec52a0035377f07540b9485884d10b34c2419411a4a5d711c4b65","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-10T01:15:20.406908Z","signature_b64":"vBsAliq4Qogd0DIS8h8n2aaDQM7rb84KYwAl30+lnpF1P3h09XRiaDvVQIViK2CQB+37rZ+WETcpxcY5yEQeBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0da2165c0b9ec52a0035377f07540b9485884d10b34c2419411a4a5d711c4b65","last_reissued_at":"2026-08-10T01:15:20.404569Z","signature_status":"signed_v1","first_computed_at":"2026-08-10T01:15:20.404569Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2608.07419","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":"2026-08-10T01:15:20Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"qIWeneUsjh7u2umSxtBQAYwGNXHEc/NhEkA41PX36NHDK2FXZ66W9UIfl1DqMUL/rknCCANUlZCzZJRKPGnzBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-15T22:05:57.731141Z"},"content_sha256":"e2e229041a72f974a9fd831a265bd146f2b1a14a546ef94abb2496b0fb83ab03","schema_version":"1.0","event_id":"sha256:e2e229041a72f974a9fd831a265bd146f2b1a14a546ef94abb2496b0fb83ab03"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:BWRBMXALT3CSUABVG57QOVALSS","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Beyond Post-Hoc Temperature Scaling: Bilevel Optimization for LLM Calibration","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Bojian Hou, Jiancong Xiao, Ruochen Jin, Zhanliang Wang, Zongyu Dai","submitted_at":"2026-08-07T17:05:10Z","abstract_excerpt":"Preference alignment often makes large language models (LLMs) overconfident and poorly calibrated. Traditional post-hoc temperature scaling is inherently domain-dependent: a temperature fitted on one domain does not generalize across domains. This motivates us to modify model parameters during training to improve calibration. We propose maximizing the entropy of predictive distributions as the calibration objective, which directly targets overconfidence by discouraging overly concentrated predictions. Inspired by temperature scaling, we realize this through a bilevel optimization formulation, "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.07419","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/2608.07419/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":"2026-08-10T01:15:20Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Mz6Lis9jagkoKLiwykeWGkDi2GkiMXEgj582nrux7+SRVdMHaQpcFRLcJcJZVc1cwySwXfbRLy/NJu91P8ZkCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-15T22:05:57.732140Z"},"content_sha256":"5c41b16b547d09aa291afe71de67df7b575bf565f92760b08aee1dfc3d9d0952","schema_version":"1.0","event_id":"sha256:5c41b16b547d09aa291afe71de67df7b575bf565f92760b08aee1dfc3d9d0952"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/BWRBMXALT3CSUABVG57QOVALSS/bundle.json","state_url":"https://pith.science/pith/BWRBMXALT3CSUABVG57QOVALSS/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/BWRBMXALT3CSUABVG57QOVALSS/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":"2026-08-15T22:05:57Z","links":{"resolver":"https://pith.science/pith/BWRBMXALT3CSUABVG57QOVALSS","bundle":"https://pith.science/pith/BWRBMXALT3CSUABVG57QOVALSS/bundle.json","state":"https://pith.science/pith/BWRBMXALT3CSUABVG57QOVALSS/state.json","well_known_bundle":"https://pith.science/.well-known/pith/BWRBMXALT3CSUABVG57QOVALSS/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:BWRBMXALT3CSUABVG57QOVALSS","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":"19a2301de88efa8b0d9f90ea54068aa63e3c88f938ffe28a36c4b0b478002bad","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-08-07T17:05:10Z","title_canon_sha256":"20d175ba28e2316001849ebbbf4d0d62ad89727c637d3ec20704fe51286be963"},"schema_version":"1.0","source":{"id":"2608.07419","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2608.07419","created_at":"2026-08-10T01:15:20Z"},{"alias_kind":"arxiv_version","alias_value":"2608.07419v1","created_at":"2026-08-10T01:15:20Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.07419","created_at":"2026-08-10T01:15:20Z"},{"alias_kind":"pith_short_12","alias_value":"BWRBMXALT3CS","created_at":"2026-08-10T01:15:20Z"},{"alias_kind":"pith_short_16","alias_value":"BWRBMXALT3CSUABV","created_at":"2026-08-10T01:15:20Z"},{"alias_kind":"pith_short_8","alias_value":"BWRBMXAL","created_at":"2026-08-10T01:15:20Z"}],"graph_snapshots":[{"event_id":"sha256:5c41b16b547d09aa291afe71de67df7b575bf565f92760b08aee1dfc3d9d0952","target":"graph","created_at":"2026-08-10T01:15:20Z","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/2608.07419/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Preference alignment often makes large language models (LLMs) overconfident and poorly calibrated. Traditional post-hoc temperature scaling is inherently domain-dependent: a temperature fitted on one domain does not generalize across domains. This motivates us to modify model parameters during training to improve calibration. We propose maximizing the entropy of predictive distributions as the calibration objective, which directly targets overconfidence by discouraging overly concentrated predictions. Inspired by temperature scaling, we realize this through a bilevel optimization formulation, ","authors_text":"Bojian Hou, Jiancong Xiao, Ruochen Jin, Zhanliang Wang, Zongyu Dai","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-08-07T17:05:10Z","title":"Beyond Post-Hoc Temperature Scaling: Bilevel Optimization for LLM Calibration"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.07419","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:e2e229041a72f974a9fd831a265bd146f2b1a14a546ef94abb2496b0fb83ab03","target":"record","created_at":"2026-08-10T01:15:20Z","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":"19a2301de88efa8b0d9f90ea54068aa63e3c88f938ffe28a36c4b0b478002bad","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-08-07T17:05:10Z","title_canon_sha256":"20d175ba28e2316001849ebbbf4d0d62ad89727c637d3ec20704fe51286be963"},"schema_version":"1.0","source":{"id":"2608.07419","kind":"arxiv","version":1}},"canonical_sha256":"0da2165c0b9ec52a0035377f07540b9485884d10b34c2419411a4a5d711c4b65","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"0da2165c0b9ec52a0035377f07540b9485884d10b34c2419411a4a5d711c4b65","first_computed_at":"2026-08-10T01:15:20.404569Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-08-10T01:15:20.404569Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"vBsAliq4Qogd0DIS8h8n2aaDQM7rb84KYwAl30+lnpF1P3h09XRiaDvVQIViK2CQB+37rZ+WETcpxcY5yEQeBw==","signature_status":"signed_v1","signed_at":"2026-08-10T01:15:20.406908Z","signed_message":"canonical_sha256_bytes"},"source_id":"2608.07419","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:e2e229041a72f974a9fd831a265bd146f2b1a14a546ef94abb2496b0fb83ab03","sha256:5c41b16b547d09aa291afe71de67df7b575bf565f92760b08aee1dfc3d9d0952"],"state_sha256":"301c48b6ebaa23c62675067749a4b20d8d63c3601723edf9347e5d28bbb7ee4e"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"zqmRKpSlqq+MhbCR/3UKBDDKiqxP6zIcOL/aUGTyeyF7S9qGunwuh8jGUCAm1Zc49x+UwrcHEaFYdpGOAREmAg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-15T22:05:57.737602Z","bundle_sha256":"8fcf1ff87f56c03d2353c0c718d02f073cfdac1ff920f4d64df21bd433249466"}}