{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:J2A5EGLFRJRKYDAACPYE6ABCTM","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":"10d2f7dd43887f0a044320abd6aca13826cc64a402c84c55fb30a8ee5b52eaa4","cross_cats_sorted":["cs.AI","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-12T16:41:09Z","title_canon_sha256":"078b5893e734b337f8900d18e5057ee95382897a63ea6c62ef0eb98bba77b3e1"},"schema_version":"1.0","source":{"id":"2605.12374","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.12374","created_at":"2026-05-20T00:00:43Z"},{"alias_kind":"arxiv_version","alias_value":"2605.12374v2","created_at":"2026-05-20T00:00:43Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.12374","created_at":"2026-05-20T00:00:43Z"},{"alias_kind":"pith_short_12","alias_value":"J2A5EGLFRJRK","created_at":"2026-05-20T00:00:43Z"},{"alias_kind":"pith_short_16","alias_value":"J2A5EGLFRJRKYDAA","created_at":"2026-05-20T00:00:43Z"},{"alias_kind":"pith_short_8","alias_value":"J2A5EGLF","created_at":"2026-05-20T00:00:43Z"}],"graph_snapshots":[{"event_id":"sha256:a75655b51b725bb17b6789bef628b88886412b89abd2d024c649fff46e5e1008","target":"graph","created_at":"2026-05-20T00:00:43Z","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":"On Qwen2.5-VL 7B, the resulting model achieves the best mean aggregate perception and reasoning performance among our supervised variants. Inference-time intervention probing further suggests that generated latents provide task-relevant visual signal beyond merely adding token slots."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The feature-space mismatch between decoder hidden states and input embeddings in pre-norm MLLMs is a primary contributor to instability in existing output-as-input visual-latent methods."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"GAP introduces three-level alignment for visual latent reasoning in MLLMs, achieving top aggregate perception and reasoning performance on Qwen2.5-VL 7B by addressing decoder-input norm mismatch."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Granular alignment at three levels lets MLLMs generate stable visual latents by fixing decoder-to-input mismatch."}],"snapshot_sha256":"312c089eecf315deb59a4b908c7c1d72903862874c4b1a7e7c45daffb66607c7"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"a5f1b36fc9fe7c5839b360e70d1c355e19646d8baa9ffe851ed75936ef692891"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-19T22:41:58.233759Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T10:37:07.135846Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T08:01:18.885772Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T07:36:34.939531Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.12374/integrity.json","findings":[],"snapshot_sha256":"7ca5a5ba5f26fc82ef388b31edef12f3b35ef689167d37d2524c2adb8416869b","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Visual latent reasoning lets a multimodal large language model (MLLM) create intermediate visual evidence as continuous tokens, avoiding external tools or image generators. However, existing methods usually follow an output-as-input latent paradigm and yield unstable gains. We identify evidence for a feature-space mismatch that can contribute to this instability: dominant visual-latent models build on pre-norm MLLMs and reuse decoder hidden states as predicted latent inputs, even though these states occupy a substantially different norm regime from the input embeddings the model was trained to","authors_text":"Dexin Wang, Guanjun Jiang, Hao Li, Lei Lv, Li Wang, Mengyu Zhou, Pascal Poupart, Qi Zhao, Xiaoxi Jiang, Yanting Miao, Yutao Sun","cross_cats":["cs.AI","cs.LG"],"headline":"Granular alignment at three levels lets MLLMs generate stable visual latents by fixing decoder-to-input mismatch.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-12T16:41:09Z","title":"Fill the GAP: A Granular Alignment Paradigm for Visual Reasoning in Multimodal Large Language Models"},"references":{"count":30,"internal_anchors":16,"resolved_work":30,"sample":[{"cited_arxiv_id":"2501.07542","doi":"","is_internal_anchor":true,"ref_index":1,"title":"Imagine while Reasoning in Space: Multimodal Visualization-of-Thought","work_id":"fe549c4a-19fa-421a-9a30-230932ba737e","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"arXiv preprint arXiv:2510.24514 , year=","work_id":"b88a3f18-3bbc-44da-82ee-292b974b146d","year":null},{"cited_arxiv_id":"2503.19786","doi":"","is_internal_anchor":true,"ref_index":3,"title":"Gemma 3 Technical Report","work_id":"f93e08bf-9e96-409b-8ac6-b8385fd17fd7","year":null},{"cited_arxiv_id":"2601.03267","doi":"","is_internal_anchor":true,"ref_index":4,"title":"OpenAI GPT-5 System Card","work_id":"ca87689a-0d29-4476-b504-b65dbbb08af4","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Qwen3-VL Technical Report , author=. 2025 , eprint=","work_id":"a6efb89e-81e1-4208-bbf7-b8686a399f3b","year":2025}],"snapshot_sha256":"b9f34aa8da27cf1d3e7850ed1bb1604d2ec924fe046279650c06afb37704faef"},"source":{"id":"2605.12374","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-19T16:43:15.588650Z","id":"02fb4ab2-0e36-4c4a-ace2-4ee9540cf8e4","model_set":{"reader":"grok-4.3"},"one_line_summary":"GAP introduces three-level alignment for visual latent reasoning in MLLMs, achieving top aggregate perception and reasoning performance on Qwen2.5-VL 7B by addressing decoder-input norm mismatch.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Granular alignment at three levels lets MLLMs generate stable visual latents by fixing decoder-to-input mismatch.","strongest_claim":"On Qwen2.5-VL 7B, the resulting model achieves the best mean aggregate perception and reasoning performance among our supervised variants. Inference-time intervention probing further suggests that generated latents provide task-relevant visual signal beyond merely adding token slots.","weakest_assumption":"The feature-space mismatch between decoder hidden states and input embeddings in pre-norm MLLMs is a primary contributor to instability in existing output-as-input visual-latent methods."}},"verdict_id":"02fb4ab2-0e36-4c4a-ace2-4ee9540cf8e4"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:bb3e1fa9b37fc90836a538df0ea73b8394c3c416192c9e0769b77c3415717c7d","target":"record","created_at":"2026-05-20T00:00:43Z","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":"10d2f7dd43887f0a044320abd6aca13826cc64a402c84c55fb30a8ee5b52eaa4","cross_cats_sorted":["cs.AI","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-12T16:41:09Z","title_canon_sha256":"078b5893e734b337f8900d18e5057ee95382897a63ea6c62ef0eb98bba77b3e1"},"schema_version":"1.0","source":{"id":"2605.12374","kind":"arxiv","version":2}},"canonical_sha256":"4e81d219658a62ac0c0013f04f00229b27cb8e84219c055bab85f053938dad0f","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"4e81d219658a62ac0c0013f04f00229b27cb8e84219c055bab85f053938dad0f","first_computed_at":"2026-05-20T00:00:43.191487Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:00:43.191487Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"HUYT+KTfXAzT+QZ3EElqzn/huf+JSdDEYGwOfzm9YqLv29gf2+8Lh9BKseNXA8LERT14C6vGxG5LSNP9rinBBA==","signature_status":"signed_v1","signed_at":"2026-05-20T00:00:43.192371Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.12374","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:bb3e1fa9b37fc90836a538df0ea73b8394c3c416192c9e0769b77c3415717c7d","sha256:a75655b51b725bb17b6789bef628b88886412b89abd2d024c649fff46e5e1008"],"state_sha256":"0a932dfe8476dcf3c59078372536a2f4ce5feeb0f7f2dd37b4eb298861cef29b"}