{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:O76FD3Q3JCZJU2MQ6W7I7YPPO2","short_pith_number":"pith:O76FD3Q3","schema_version":"1.0","canonical_sha256":"77fc51ee1b48b29a6990f5be8fe1ef76890e2c1ea8ad3728dbfe6fc7508d563e","source":{"kind":"arxiv","id":"2605.29951","version":1},"attestation_state":"computed","paper":{"title":"MuPHI: Learning Implicit Multimodal Harm Reasoning via Semantically Grounded Reward Optimization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.LG","cs.MM"],"primary_cat":"cs.AI","authors_text":"Anisha Saha, Sophia Wiedmann, Teodora Kamova, Timothy Hospedales, Varsha Suresh, Vera Demberg","submitted_at":"2026-05-28T13:58:15Z","abstract_excerpt":"Understanding how harm emerges from interaction between otherwise benign image-text pairs requires intent-aware cross-modal reasoning beyond surface-level features. Existing vision-language models (VLMs) excel at literal reasoning over perceptual cues but often fail to derive harmful semantics that rely on implicit, context-dependent reasoning. To evaluate VLMs on compositional harm detection and reasoning, we introduce Multimodal Pragmatic Harm Interpretation (MuPHI), a dataset containing image-text pairs where harm is encoded in subtle multimodal cues. MuPHI spans diverse harm categories and"},"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":"2605.29951","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-05-28T13:58:15Z","cross_cats_sorted":["cs.CL","cs.LG","cs.MM"],"title_canon_sha256":"c9b774c58aa52396a13c06fdec6f8716184f92a9c9b2927786b8886e0652d529","abstract_canon_sha256":"9781b0f8de28bcd53b28abedd6ec549627ca0290d9603a430c4a6d8527cb0c41"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-29T02:06:03.401644Z","signature_b64":"cgI6Sel5q0ygb7KWyXjOylp7rxi0u4/vdL6snIBGW29hinFeqOBvv8Y+RQ4FfulY00+EZR90gWxoCuoH6x5pBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"77fc51ee1b48b29a6990f5be8fe1ef76890e2c1ea8ad3728dbfe6fc7508d563e","last_reissued_at":"2026-05-29T02:06:03.400820Z","signature_status":"signed_v1","first_computed_at":"2026-05-29T02:06:03.400820Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MuPHI: Learning Implicit Multimodal Harm Reasoning via Semantically Grounded Reward Optimization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.LG","cs.MM"],"primary_cat":"cs.AI","authors_text":"Anisha Saha, Sophia Wiedmann, Teodora Kamova, Timothy Hospedales, Varsha Suresh, Vera Demberg","submitted_at":"2026-05-28T13:58:15Z","abstract_excerpt":"Understanding how harm emerges from interaction between otherwise benign image-text pairs requires intent-aware cross-modal reasoning beyond surface-level features. Existing vision-language models (VLMs) excel at literal reasoning over perceptual cues but often fail to derive harmful semantics that rely on implicit, context-dependent reasoning. To evaluate VLMs on compositional harm detection and reasoning, we introduce Multimodal Pragmatic Harm Interpretation (MuPHI), a dataset containing image-text pairs where harm is encoded in subtle multimodal cues. MuPHI spans diverse harm categories and"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.29951","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/2605.29951/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":"2605.29951","created_at":"2026-05-29T02:06:03.400952+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.29951v1","created_at":"2026-05-29T02:06:03.400952+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.29951","created_at":"2026-05-29T02:06:03.400952+00:00"},{"alias_kind":"pith_short_12","alias_value":"O76FD3Q3JCZJ","created_at":"2026-05-29T02:06:03.400952+00:00"},{"alias_kind":"pith_short_16","alias_value":"O76FD3Q3JCZJU2MQ","created_at":"2026-05-29T02:06:03.400952+00:00"},{"alias_kind":"pith_short_8","alias_value":"O76FD3Q3","created_at":"2026-05-29T02:06:03.400952+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/O76FD3Q3JCZJU2MQ6W7I7YPPO2","json":"https://pith.science/pith/O76FD3Q3JCZJU2MQ6W7I7YPPO2.json","graph_json":"https://pith.science/api/pith-number/O76FD3Q3JCZJU2MQ6W7I7YPPO2/graph.json","events_json":"https://pith.science/api/pith-number/O76FD3Q3JCZJU2MQ6W7I7YPPO2/events.json","paper":"https://pith.science/paper/O76FD3Q3"},"agent_actions":{"view_html":"https://pith.science/pith/O76FD3Q3JCZJU2MQ6W7I7YPPO2","download_json":"https://pith.science/pith/O76FD3Q3JCZJU2MQ6W7I7YPPO2.json","view_paper":"https://pith.science/paper/O76FD3Q3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.29951&json=true","fetch_graph":"https://pith.science/api/pith-number/O76FD3Q3JCZJU2MQ6W7I7YPPO2/graph.json","fetch_events":"https://pith.science/api/pith-number/O76FD3Q3JCZJU2MQ6W7I7YPPO2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/O76FD3Q3JCZJU2MQ6W7I7YPPO2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/O76FD3Q3JCZJU2MQ6W7I7YPPO2/action/storage_attestation","attest_author":"https://pith.science/pith/O76FD3Q3JCZJU2MQ6W7I7YPPO2/action/author_attestation","sign_citation":"https://pith.science/pith/O76FD3Q3JCZJU2MQ6W7I7YPPO2/action/citation_signature","submit_replication":"https://pith.science/pith/O76FD3Q3JCZJU2MQ6W7I7YPPO2/action/replication_record"}},"created_at":"2026-05-29T02:06:03.400952+00:00","updated_at":"2026-05-29T02:06:03.400952+00:00"}