{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:JBZOYWTEQZRAQS4IOP4MBCKJ5S","short_pith_number":"pith:JBZOYWTE","canonical_record":{"source":{"id":"2605.12208","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2026-05-12T14:46:08Z","cross_cats_sorted":["cs.AI","cs.LG","stat.CO"],"title_canon_sha256":"dbf3db7ed915a678d929033294cc1d6aca7904d99458f3091ad107b6898f7aae","abstract_canon_sha256":"9fe93b5157c2d1916dc32ae588993288bb10e66dae823241734d270dea902d9f"},"schema_version":"1.0"},"canonical_sha256":"4872ec5a648662084b8873f8c08949ec975aef00bb27c523bfed9896560f7e2d","source":{"kind":"arxiv","id":"2605.12208","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.12208","created_at":"2026-05-29T01:05:12Z"},{"alias_kind":"arxiv_version","alias_value":"2605.12208v2","created_at":"2026-05-29T01:05:12Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.12208","created_at":"2026-05-29T01:05:12Z"},{"alias_kind":"pith_short_12","alias_value":"JBZOYWTEQZRA","created_at":"2026-05-29T01:05:12Z"},{"alias_kind":"pith_short_16","alias_value":"JBZOYWTEQZRAQS4I","created_at":"2026-05-29T01:05:12Z"},{"alias_kind":"pith_short_8","alias_value":"JBZOYWTE","created_at":"2026-05-29T01:05:12Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:JBZOYWTEQZRAQS4IOP4MBCKJ5S","target":"record","payload":{"canonical_record":{"source":{"id":"2605.12208","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2026-05-12T14:46:08Z","cross_cats_sorted":["cs.AI","cs.LG","stat.CO"],"title_canon_sha256":"dbf3db7ed915a678d929033294cc1d6aca7904d99458f3091ad107b6898f7aae","abstract_canon_sha256":"9fe93b5157c2d1916dc32ae588993288bb10e66dae823241734d270dea902d9f"},"schema_version":"1.0"},"canonical_sha256":"4872ec5a648662084b8873f8c08949ec975aef00bb27c523bfed9896560f7e2d","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-29T01:05:12.308447Z","signature_b64":"yolG+s+UBrQcdkf3xbDLh/ldKZBXwEg0C56PHZmrGaYHaFafV6rY9Nmuq3m1E2BN7+z0P1hugmO4uTXDRJ/OAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4872ec5a648662084b8873f8c08949ec975aef00bb27c523bfed9896560f7e2d","last_reissued_at":"2026-05-29T01:05:12.307864Z","signature_status":"signed_v1","first_computed_at":"2026-05-29T01:05:12.307864Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.12208","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-29T01:05:12Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"DTIDHcN3XS6FcZ+DRrnTBKdCkpsZ20LMLJ6gc46TC8MC4C0g56B5f3LIG/aLyn7aMnx07iNiBC1WwQbvqwpyBA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-29T07:17:21.080578Z"},"content_sha256":"c8f43861593a1d5bd0cccbcf450a761c1c099db4d51e8731b34ab29ebc22ab1b","schema_version":"1.0","event_id":"sha256:c8f43861593a1d5bd0cccbcf450a761c1c099db4d51e8731b34ab29ebc22ab1b"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:JBZOYWTEQZRAQS4IOP4MBCKJ5S","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Self-Supervised Laplace Approximation for Bayesian Uncertainty Quantification","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Refitting models on their own predictions approximates the posterior predictive distribution directly and improves calibration over classical Laplace methods.","cross_cats":["cs.AI","cs.LG","stat.CO"],"primary_cat":"stat.ML","authors_text":"Alexander Marquard, Julian Rodemann, Michele Caprio, Thomas Augustin","submitted_at":"2026-05-12T14:46:08Z","abstract_excerpt":"Approximate Bayesian inference typically revolves around computing the posterior parameter distribution. In practice, however, the main object of interest is often a model's predictions rather than its parameters. In this work, we propose to bypass the parameter posterior and focus directly on approximating the posterior predictive distribution. We achieve this by drawing inspiration from self-training within self-supervised and semi-supervised learning. Essentially, we quantify a Bayesian model's predictive uncertainty by refitting on self-predicted data. The idea is strikingly simple: If a m"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Across a wide array of regression tasks with simulated and real-world datasets, our methods outperform classical Laplace approximations in predictive calibration while remaining computationally efficient.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That refitting the model on self-predicted data effectively approximates the posterior predictive distribution, assuming the initial model predictions are sufficiently reliable to serve as pseudo-labels for uncertainty quantification.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SSLA approximates the posterior predictive distribution by refitting Bayesian models on self-predicted data, providing a sampling-free method that improves predictive calibration over classical Laplace approximations in regression tasks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Refitting models on their own predictions approximates the posterior predictive distribution directly and improves calibration over classical Laplace methods.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"7bfd5b1c3d3d84fd7b8d5712e51e01aa086f265f01a94d099eeb8bd078e04e03"},"source":{"id":"2605.12208","kind":"arxiv","version":2},"verdict":{"id":"d9e8d456-4160-49a9-8c41-9e25657dcd45","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-13T04:44:02.157733Z","strongest_claim":"Across a wide array of regression tasks with simulated and real-world datasets, our methods outperform classical Laplace approximations in predictive calibration while remaining computationally efficient.","one_line_summary":"SSLA approximates the posterior predictive distribution by refitting Bayesian models on self-predicted data, providing a sampling-free method that improves predictive calibration over classical Laplace approximations in regression tasks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That refitting the model on self-predicted data effectively approximates the posterior predictive distribution, assuming the initial model predictions are sufficiently reliable to serve as pseudo-labels for uncertainty quantification.","pith_extraction_headline":"Refitting models on their own predictions approximates the posterior predictive distribution directly and improves calibration over classical Laplace methods."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.12208/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-26T14:46:39.938422Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-20T14:31:25.554712Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-20T10:35:24.016190Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T22:41:58.374605Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"a94faf8b249891b16faf601f92ffb1027431819a581ca6ac28c722efa2ac2de0"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"cd9383e3706a65b6077cdb4e45f67ef0c8a3fa5242372efba71c1af676f7e068"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"d9e8d456-4160-49a9-8c41-9e25657dcd45"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-29T01:05:12Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"XftOgMBxUmDxc0ZXziNwJGX+3CLGbWKP8cwiYByrEhoZUJcNCCM91f+9Xb6hgUQ5HzQyiDDs1/asPOvi7l7kDQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-29T07:17:21.081106Z"},"content_sha256":"953b76001a4c5555ac8d16d803caed2b1ff8cea9b83f2be5e2a5d1d284c3a7ed","schema_version":"1.0","event_id":"sha256:953b76001a4c5555ac8d16d803caed2b1ff8cea9b83f2be5e2a5d1d284c3a7ed"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/JBZOYWTEQZRAQS4IOP4MBCKJ5S/bundle.json","state_url":"https://pith.science/pith/JBZOYWTEQZRAQS4IOP4MBCKJ5S/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/JBZOYWTEQZRAQS4IOP4MBCKJ5S/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":"2026-06-29T07:17:21Z","links":{"resolver":"https://pith.science/pith/JBZOYWTEQZRAQS4IOP4MBCKJ5S","bundle":"https://pith.science/pith/JBZOYWTEQZRAQS4IOP4MBCKJ5S/bundle.json","state":"https://pith.science/pith/JBZOYWTEQZRAQS4IOP4MBCKJ5S/state.json","well_known_bundle":"https://pith.science/.well-known/pith/JBZOYWTEQZRAQS4IOP4MBCKJ5S/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:JBZOYWTEQZRAQS4IOP4MBCKJ5S","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":"9fe93b5157c2d1916dc32ae588993288bb10e66dae823241734d270dea902d9f","cross_cats_sorted":["cs.AI","cs.LG","stat.CO"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2026-05-12T14:46:08Z","title_canon_sha256":"dbf3db7ed915a678d929033294cc1d6aca7904d99458f3091ad107b6898f7aae"},"schema_version":"1.0","source":{"id":"2605.12208","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.12208","created_at":"2026-05-29T01:05:12Z"},{"alias_kind":"arxiv_version","alias_value":"2605.12208v2","created_at":"2026-05-29T01:05:12Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.12208","created_at":"2026-05-29T01:05:12Z"},{"alias_kind":"pith_short_12","alias_value":"JBZOYWTEQZRA","created_at":"2026-05-29T01:05:12Z"},{"alias_kind":"pith_short_16","alias_value":"JBZOYWTEQZRAQS4I","created_at":"2026-05-29T01:05:12Z"},{"alias_kind":"pith_short_8","alias_value":"JBZOYWTE","created_at":"2026-05-29T01:05:12Z"}],"graph_snapshots":[{"event_id":"sha256:953b76001a4c5555ac8d16d803caed2b1ff8cea9b83f2be5e2a5d1d284c3a7ed","target":"graph","created_at":"2026-05-29T01:05:12Z","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":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Across a wide array of regression tasks with simulated and real-world datasets, our methods outperform classical Laplace approximations in predictive calibration while remaining computationally efficient."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That refitting the model on self-predicted data effectively approximates the posterior predictive distribution, assuming the initial model predictions are sufficiently reliable to serve as pseudo-labels for uncertainty quantification."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"SSLA approximates the posterior predictive distribution by refitting Bayesian models on self-predicted data, providing a sampling-free method that improves predictive calibration over classical Laplace approximations in regression tasks."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Refitting models on their own predictions approximates the posterior predictive distribution directly and improves calibration over classical Laplace methods."}],"snapshot_sha256":"7bfd5b1c3d3d84fd7b8d5712e51e01aa086f265f01a94d099eeb8bd078e04e03"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"cd9383e3706a65b6077cdb4e45f67ef0c8a3fa5242372efba71c1af676f7e068"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-26T14:46:39.938422Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-20T14:31:25.554712Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-20T10:35:24.016190Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-19T22:41:58.374605Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.12208/integrity.json","findings":[],"snapshot_sha256":"a94faf8b249891b16faf601f92ffb1027431819a581ca6ac28c722efa2ac2de0","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Approximate Bayesian inference typically revolves around computing the posterior parameter distribution. In practice, however, the main object of interest is often a model's predictions rather than its parameters. In this work, we propose to bypass the parameter posterior and focus directly on approximating the posterior predictive distribution. We achieve this by drawing inspiration from self-training within self-supervised and semi-supervised learning. Essentially, we quantify a Bayesian model's predictive uncertainty by refitting on self-predicted data. The idea is strikingly simple: If a m","authors_text":"Alexander Marquard, Julian Rodemann, Michele Caprio, Thomas Augustin","cross_cats":["cs.AI","cs.LG","stat.CO"],"headline":"Refitting models on their own predictions approximates the posterior predictive distribution directly and improves calibration over classical Laplace methods.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2026-05-12T14:46:08Z","title":"Self-Supervised Laplace Approximation for Bayesian Uncertainty Quantification"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.12208","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-13T04:44:02.157733Z","id":"d9e8d456-4160-49a9-8c41-9e25657dcd45","model_set":{"reader":"grok-4.3"},"one_line_summary":"SSLA approximates the posterior predictive distribution by refitting Bayesian models on self-predicted data, providing a sampling-free method that improves predictive calibration over classical Laplace approximations in regression tasks.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Refitting models on their own predictions approximates the posterior predictive distribution directly and improves calibration over classical Laplace methods.","strongest_claim":"Across a wide array of regression tasks with simulated and real-world datasets, our methods outperform classical Laplace approximations in predictive calibration while remaining computationally efficient.","weakest_assumption":"That refitting the model on self-predicted data effectively approximates the posterior predictive distribution, assuming the initial model predictions are sufficiently reliable to serve as pseudo-labels for uncertainty quantification."}},"verdict_id":"d9e8d456-4160-49a9-8c41-9e25657dcd45"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:c8f43861593a1d5bd0cccbcf450a761c1c099db4d51e8731b34ab29ebc22ab1b","target":"record","created_at":"2026-05-29T01:05:12Z","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":"9fe93b5157c2d1916dc32ae588993288bb10e66dae823241734d270dea902d9f","cross_cats_sorted":["cs.AI","cs.LG","stat.CO"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2026-05-12T14:46:08Z","title_canon_sha256":"dbf3db7ed915a678d929033294cc1d6aca7904d99458f3091ad107b6898f7aae"},"schema_version":"1.0","source":{"id":"2605.12208","kind":"arxiv","version":2}},"canonical_sha256":"4872ec5a648662084b8873f8c08949ec975aef00bb27c523bfed9896560f7e2d","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"4872ec5a648662084b8873f8c08949ec975aef00bb27c523bfed9896560f7e2d","first_computed_at":"2026-05-29T01:05:12.307864Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-29T01:05:12.307864Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"yolG+s+UBrQcdkf3xbDLh/ldKZBXwEg0C56PHZmrGaYHaFafV6rY9Nmuq3m1E2BN7+z0P1hugmO4uTXDRJ/OAA==","signature_status":"signed_v1","signed_at":"2026-05-29T01:05:12.308447Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.12208","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:c8f43861593a1d5bd0cccbcf450a761c1c099db4d51e8731b34ab29ebc22ab1b","sha256:953b76001a4c5555ac8d16d803caed2b1ff8cea9b83f2be5e2a5d1d284c3a7ed"],"state_sha256":"3c4eb622c953d3a3f771fea42a9f4b744830fcbdd298b8119416e7f09c0683e9"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"a3mmuy4QDQpOjURUb6JbalQ8IZq/r3YaCiqxkBAUgYPOkGEv8DcKFZJve1kSVeY6bJIw6EVY9QuPXr9fe5gICQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-29T07:17:21.085103Z","bundle_sha256":"7a1d0362520889bafa144cdfdb6f4cb3a673b2530a711616f201de9195fe05ff"}}