{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:PSKIOVTH3OSE3VZCPTMMBIBFOW","short_pith_number":"pith:PSKIOVTH","schema_version":"1.0","canonical_sha256":"7c94875667dba44dd7227cd8c0a0257580243f89bf2e37c6043c5ed0a04e4bde","source":{"kind":"arxiv","id":"2604.16358","version":2},"attestation_state":"computed","paper":{"title":"SaFeR-Steer: Evolving Multi-Turn MLLMs via Synthetic Bootstrapping and Feedback Dynamics","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"SaFeR-Steer trains multi-turn multimodal models to hold safety and helpfulness against escalating attacks through synthetic bootstrapping and tutor feedback.","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"An Zhang, Hanyu Li, Haolong Hu, Huahui Yi, Kun Wang, Qiankun Li, Tiancheng He, Yang Liu, Zhigang Zeng","submitted_at":"2026-03-18T09:28:29Z","abstract_excerpt":"MLLMs are increasingly deployed in multi-turn settings, where attackers can escalate unsafe intent through the evolving visual-text history and exploit long-context safety decay. Yet safety alignment is still dominated by single-turn data and fixed-template dialogues, leaving a mismatch between training and deployment.\n  To bridge this gap, we propose SaFeR-Steer, a progressive multi-turn alignment framework that combines staged synthetic bootstrapping with tutor-in-the-loop GRPO to train a single student under adaptive, on-policy attacks. We also introduce Trajectory-Consistent Summative Rewa"},"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":"2604.16358","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-03-18T09:28:29Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"72f46ab6d5f941f24e43ae73fcf421145ce251a8b0b3013c6b065c1aabaaeefa","abstract_canon_sha256":"01084c11191a11914693f030ba9811d9905f01d853b535a544793ac371bc6d3c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T02:04:47.904341Z","signature_b64":"nCUjOkuGECr3j0M1bMIse5cYOjNBx3geVGkTQ9paZqKPc6s82qZL9eM0jJhrdZmdRrt2OONVuDdMhf2Te80eBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7c94875667dba44dd7227cd8c0a0257580243f89bf2e37c6043c5ed0a04e4bde","last_reissued_at":"2026-05-28T02:04:47.903904Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T02:04:47.903904Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SaFeR-Steer: Evolving Multi-Turn MLLMs via Synthetic Bootstrapping and Feedback Dynamics","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"SaFeR-Steer trains multi-turn multimodal models to hold safety and helpfulness against escalating attacks through synthetic bootstrapping and tutor feedback.","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"An Zhang, Hanyu Li, Haolong Hu, Huahui Yi, Kun Wang, Qiankun Li, Tiancheng He, Yang Liu, Zhigang Zeng","submitted_at":"2026-03-18T09:28:29Z","abstract_excerpt":"MLLMs are increasingly deployed in multi-turn settings, where attackers can escalate unsafe intent through the evolving visual-text history and exploit long-context safety decay. Yet safety alignment is still dominated by single-turn data and fixed-template dialogues, leaving a mismatch between training and deployment.\n  To bridge this gap, we propose SaFeR-Steer, a progressive multi-turn alignment framework that combines staged synthetic bootstrapping with tutor-in-the-loop GRPO to train a single student under adaptive, on-policy attacks. We also introduce Trajectory-Consistent Summative Rewa"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Starting from Qwen2.5-VL-3B/7B, SaFeR-Steer substantially improves Safety/Helpfulness on both single-turn (48.30/45.86 -> 81.84/70.77 for 3B; 56.21/60.32 -> 87.89/77.40 for 7B) and multi-turn benchmarks (12.55/27.13 -> 55.58/70.27 for 3B; 24.66/46.48 -> 64.89/72.35 for 7B), shifting failures to later turns.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the synthetic bootstrapping process and tutor-in-the-loop GRPO produce data and gradients that generalize to real-world multi-turn attacks rather than overfitting to the generated distribution or the specific tutor model.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SaFeR-Steer boosts single-turn and multi-turn safety/helpfulness scores in Qwen2.5-VL models via staged synthetic bootstrapping and tutor-in-the-loop GRPO on a new STEER dataset.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"SaFeR-Steer trains multi-turn multimodal models to hold safety and helpfulness against escalating attacks through synthetic bootstrapping and tutor feedback.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"eeaf42adeb64867baf3def447b50c953ac971f612f45c14d63086572416a3fe8"},"source":{"id":"2604.16358","kind":"arxiv","version":2},"verdict":{"id":"e58ee4d9-d3c1-4f7d-be42-df15021160bf","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T09:46:57.358685Z","strongest_claim":"Starting from Qwen2.5-VL-3B/7B, SaFeR-Steer substantially improves Safety/Helpfulness on both single-turn (48.30/45.86 -> 81.84/70.77 for 3B; 56.21/60.32 -> 87.89/77.40 for 7B) and multi-turn benchmarks (12.55/27.13 -> 55.58/70.27 for 3B; 24.66/46.48 -> 64.89/72.35 for 7B), shifting failures to later turns.","one_line_summary":"SaFeR-Steer boosts single-turn and multi-turn safety/helpfulness scores in Qwen2.5-VL models via staged synthetic bootstrapping and tutor-in-the-loop GRPO on a new STEER dataset.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the synthetic bootstrapping process and tutor-in-the-loop GRPO produce data and gradients that generalize to real-world multi-turn attacks rather than overfitting to the generated distribution or the specific tutor model.","pith_extraction_headline":"SaFeR-Steer trains multi-turn multimodal models to hold safety and helpfulness against escalating attacks through synthetic bootstrapping and tutor feedback."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.16358/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":"2604.16358","created_at":"2026-05-28T02:04:47.903967+00:00"},{"alias_kind":"arxiv_version","alias_value":"2604.16358v2","created_at":"2026-05-28T02:04:47.903967+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.16358","created_at":"2026-05-28T02:04:47.903967+00:00"},{"alias_kind":"pith_short_12","alias_value":"PSKIOVTH3OSE","created_at":"2026-05-28T02:04:47.903967+00:00"},{"alias_kind":"pith_short_16","alias_value":"PSKIOVTH3OSE3VZC","created_at":"2026-05-28T02:04:47.903967+00:00"},{"alias_kind":"pith_short_8","alias_value":"PSKIOVTH","created_at":"2026-05-28T02:04:47.903967+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/PSKIOVTH3OSE3VZCPTMMBIBFOW","json":"https://pith.science/pith/PSKIOVTH3OSE3VZCPTMMBIBFOW.json","graph_json":"https://pith.science/api/pith-number/PSKIOVTH3OSE3VZCPTMMBIBFOW/graph.json","events_json":"https://pith.science/api/pith-number/PSKIOVTH3OSE3VZCPTMMBIBFOW/events.json","paper":"https://pith.science/paper/PSKIOVTH"},"agent_actions":{"view_html":"https://pith.science/pith/PSKIOVTH3OSE3VZCPTMMBIBFOW","download_json":"https://pith.science/pith/PSKIOVTH3OSE3VZCPTMMBIBFOW.json","view_paper":"https://pith.science/paper/PSKIOVTH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2604.16358&json=true","fetch_graph":"https://pith.science/api/pith-number/PSKIOVTH3OSE3VZCPTMMBIBFOW/graph.json","fetch_events":"https://pith.science/api/pith-number/PSKIOVTH3OSE3VZCPTMMBIBFOW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PSKIOVTH3OSE3VZCPTMMBIBFOW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PSKIOVTH3OSE3VZCPTMMBIBFOW/action/storage_attestation","attest_author":"https://pith.science/pith/PSKIOVTH3OSE3VZCPTMMBIBFOW/action/author_attestation","sign_citation":"https://pith.science/pith/PSKIOVTH3OSE3VZCPTMMBIBFOW/action/citation_signature","submit_replication":"https://pith.science/pith/PSKIOVTH3OSE3VZCPTMMBIBFOW/action/replication_record"}},"created_at":"2026-05-28T02:04:47.903967+00:00","updated_at":"2026-05-28T02:04:47.903967+00:00"}