{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:5Q2KXVNCZPVXQQ3AVT6O5PWBVP","short_pith_number":"pith:5Q2KXVNC","schema_version":"1.0","canonical_sha256":"ec34abd5a2cbeb784360acfceebec1abc61bbdb18126fb9a91b8ca4613c9cbf1","source":{"kind":"arxiv","id":"2606.11046","version":1},"attestation_state":"computed","paper":{"title":"Does Reasoning Preserve Alignment? On the Trustworthiness of Large Reasoning Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Alvaro Velasquez, Amrit Singh Bedi, Avinash Reddy, Furong Huang, Prajakta Kini, Satya Sai Srinath Namburi GNVV, Souradip Chakraborty","submitted_at":"2026-06-09T16:14:27Z","abstract_excerpt":"Instruction-tuned LLMs are increasingly converted into reasoning models through post-training to improve multi-step task performance. This conversion is usually optimized for reasoning accuracy, without explicitly preserving the alignment behavior of the instruction-tuned model, such as safe refusal, bias avoidance, and privacy protection. We ask: does this conversion preserve alignment? We study this question through a trustworthiness audit and find that it is not behavior-preserving by default. For a systematic analysis, we compare reasoning models produced via supervised fine-tuning, RL-bas"},"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":"2606.11046","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-06-09T16:14:27Z","cross_cats_sorted":[],"title_canon_sha256":"86d04dd1bfacb5b2e239a4bb037246ff7192489fca2d2dc0db26d0e483660e7b","abstract_canon_sha256":"640eb9bb487a76eedcb3c54e507ab075043acd5d100adc15228babf4a3c3241a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-10T01:11:06.750333Z","signature_b64":"bAeUzW6L/r0VJJYWwGQu/PXzyspVOOT7rCYH++hyA8KEAvwnO3VW8VAB3enpEZiJVzDIwn/Ejx4BQrSxZTnFAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ec34abd5a2cbeb784360acfceebec1abc61bbdb18126fb9a91b8ca4613c9cbf1","last_reissued_at":"2026-06-10T01:11:06.749412Z","signature_status":"signed_v1","first_computed_at":"2026-06-10T01:11:06.749412Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Does Reasoning Preserve Alignment? On the Trustworthiness of Large Reasoning Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Alvaro Velasquez, Amrit Singh Bedi, Avinash Reddy, Furong Huang, Prajakta Kini, Satya Sai Srinath Namburi GNVV, Souradip Chakraborty","submitted_at":"2026-06-09T16:14:27Z","abstract_excerpt":"Instruction-tuned LLMs are increasingly converted into reasoning models through post-training to improve multi-step task performance. This conversion is usually optimized for reasoning accuracy, without explicitly preserving the alignment behavior of the instruction-tuned model, such as safe refusal, bias avoidance, and privacy protection. We ask: does this conversion preserve alignment? We study this question through a trustworthiness audit and find that it is not behavior-preserving by default. For a systematic analysis, we compare reasoning models produced via supervised fine-tuning, RL-bas"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.11046","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/2606.11046/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":"2606.11046","created_at":"2026-06-10T01:11:06.749566+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.11046v1","created_at":"2026-06-10T01:11:06.749566+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.11046","created_at":"2026-06-10T01:11:06.749566+00:00"},{"alias_kind":"pith_short_12","alias_value":"5Q2KXVNCZPVX","created_at":"2026-06-10T01:11:06.749566+00:00"},{"alias_kind":"pith_short_16","alias_value":"5Q2KXVNCZPVXQQ3A","created_at":"2026-06-10T01:11:06.749566+00:00"},{"alias_kind":"pith_short_8","alias_value":"5Q2KXVNC","created_at":"2026-06-10T01:11:06.749566+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5Q2KXVNCZPVXQQ3AVT6O5PWBVP","json":"https://pith.science/pith/5Q2KXVNCZPVXQQ3AVT6O5PWBVP.json","graph_json":"https://pith.science/api/pith-number/5Q2KXVNCZPVXQQ3AVT6O5PWBVP/graph.json","events_json":"https://pith.science/api/pith-number/5Q2KXVNCZPVXQQ3AVT6O5PWBVP/events.json","paper":"https://pith.science/paper/5Q2KXVNC"},"agent_actions":{"view_html":"https://pith.science/pith/5Q2KXVNCZPVXQQ3AVT6O5PWBVP","download_json":"https://pith.science/pith/5Q2KXVNCZPVXQQ3AVT6O5PWBVP.json","view_paper":"https://pith.science/paper/5Q2KXVNC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.11046&json=true","fetch_graph":"https://pith.science/api/pith-number/5Q2KXVNCZPVXQQ3AVT6O5PWBVP/graph.json","fetch_events":"https://pith.science/api/pith-number/5Q2KXVNCZPVXQQ3AVT6O5PWBVP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5Q2KXVNCZPVXQQ3AVT6O5PWBVP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5Q2KXVNCZPVXQQ3AVT6O5PWBVP/action/storage_attestation","attest_author":"https://pith.science/pith/5Q2KXVNCZPVXQQ3AVT6O5PWBVP/action/author_attestation","sign_citation":"https://pith.science/pith/5Q2KXVNCZPVXQQ3AVT6O5PWBVP/action/citation_signature","submit_replication":"https://pith.science/pith/5Q2KXVNCZPVXQQ3AVT6O5PWBVP/action/replication_record"}},"created_at":"2026-06-10T01:11:06.749566+00:00","updated_at":"2026-06-10T01:11:06.749566+00:00"}