{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:5OZDLVNEHPUDD45225766OKOUY","short_pith_number":"pith:5OZDLVNE","schema_version":"1.0","canonical_sha256":"ebb235d5a43be831f3bad77fef394ea6308ec36f7bf01c0803117032261837a5","source":{"kind":"arxiv","id":"2506.23329","version":1},"attestation_state":"computed","paper":{"title":"IR3D-Bench: Evaluating Vision-Language Model Scene Understanding as Agentic Inverse Rendering","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Brandon Y. Feng, Chenxin Li, Parker Liu, Sirui Han, Wuyang Li, Yipeng Wu, Yunlong Lin, Zhengxin Li, Zhenyuan Zhang, Zhiqin Yang","submitted_at":"2025-06-29T17:02:57Z","abstract_excerpt":"Vision-language models (VLMs) excel at descriptive tasks, but whether they truly understand scenes from visual observations remains uncertain. We introduce IR3D-Bench, a benchmark challenging VLMs to demonstrate understanding through active creation rather than passive recognition. Grounded in the analysis-by-synthesis paradigm, IR3D-Bench tasks Vision-Language Agents (VLAs) with actively using programming and rendering tools to recreate the underlying 3D structure of an input image, achieving agentic inverse rendering through tool use. This \"understanding-by-creating\" approach probes the tool"},"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":"2506.23329","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-06-29T17:02:57Z","cross_cats_sorted":[],"title_canon_sha256":"2c53e506dca5d44fd1d91e6725689aa4792c8463ce312111e0bd5de19377d370","abstract_canon_sha256":"883b394715e9c03e3a88930bdfcc33688e5eadeb1edcf7d47cebe23797821ced"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:29:19.881807Z","signature_b64":"kdigriiU3+gzVcusZdk5K7noy7XhlotdHLzL/z7bXhIs9ejXgJ+sNOiFNBaLxrQYbDU3wnBXoNoJVb8ty9ZNBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ebb235d5a43be831f3bad77fef394ea6308ec36f7bf01c0803117032261837a5","last_reissued_at":"2026-07-05T11:29:19.881288Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:29:19.881288Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"IR3D-Bench: Evaluating Vision-Language Model Scene Understanding as Agentic Inverse Rendering","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Brandon Y. Feng, Chenxin Li, Parker Liu, Sirui Han, Wuyang Li, Yipeng Wu, Yunlong Lin, Zhengxin Li, Zhenyuan Zhang, Zhiqin Yang","submitted_at":"2025-06-29T17:02:57Z","abstract_excerpt":"Vision-language models (VLMs) excel at descriptive tasks, but whether they truly understand scenes from visual observations remains uncertain. We introduce IR3D-Bench, a benchmark challenging VLMs to demonstrate understanding through active creation rather than passive recognition. Grounded in the analysis-by-synthesis paradigm, IR3D-Bench tasks Vision-Language Agents (VLAs) with actively using programming and rendering tools to recreate the underlying 3D structure of an input image, achieving agentic inverse rendering through tool use. This \"understanding-by-creating\" approach probes the tool"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.23329","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/2506.23329/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":"2506.23329","created_at":"2026-07-05T11:29:19.881358+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.23329v1","created_at":"2026-07-05T11:29:19.881358+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.23329","created_at":"2026-07-05T11:29:19.881358+00:00"},{"alias_kind":"pith_short_12","alias_value":"5OZDLVNEHPUD","created_at":"2026-07-05T11:29:19.881358+00:00"},{"alias_kind":"pith_short_16","alias_value":"5OZDLVNEHPUDD452","created_at":"2026-07-05T11:29:19.881358+00:00"},{"alias_kind":"pith_short_8","alias_value":"5OZDLVNE","created_at":"2026-07-05T11:29:19.881358+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.02580","citing_title":"Thinking in Blender: Staged Executable Inverse Graphics with Vision-Language Models","ref_index":49,"is_internal_anchor":false},{"citing_arxiv_id":"2604.13035","citing_title":"SceneCritic: A Symbolic Evaluator for 3D Indoor Scene Synthesis","ref_index":27,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5OZDLVNEHPUDD45225766OKOUY","json":"https://pith.science/pith/5OZDLVNEHPUDD45225766OKOUY.json","graph_json":"https://pith.science/api/pith-number/5OZDLVNEHPUDD45225766OKOUY/graph.json","events_json":"https://pith.science/api/pith-number/5OZDLVNEHPUDD45225766OKOUY/events.json","paper":"https://pith.science/paper/5OZDLVNE"},"agent_actions":{"view_html":"https://pith.science/pith/5OZDLVNEHPUDD45225766OKOUY","download_json":"https://pith.science/pith/5OZDLVNEHPUDD45225766OKOUY.json","view_paper":"https://pith.science/paper/5OZDLVNE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.23329&json=true","fetch_graph":"https://pith.science/api/pith-number/5OZDLVNEHPUDD45225766OKOUY/graph.json","fetch_events":"https://pith.science/api/pith-number/5OZDLVNEHPUDD45225766OKOUY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5OZDLVNEHPUDD45225766OKOUY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5OZDLVNEHPUDD45225766OKOUY/action/storage_attestation","attest_author":"https://pith.science/pith/5OZDLVNEHPUDD45225766OKOUY/action/author_attestation","sign_citation":"https://pith.science/pith/5OZDLVNEHPUDD45225766OKOUY/action/citation_signature","submit_replication":"https://pith.science/pith/5OZDLVNEHPUDD45225766OKOUY/action/replication_record"}},"created_at":"2026-07-05T11:29:19.881358+00:00","updated_at":"2026-07-05T11:29:19.881358+00:00"}