{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:4DXDHE5HM5MT6BNUFFLHDQ7XZJ","short_pith_number":"pith:4DXDHE5H","canonical_record":{"source":{"id":"2403.18814","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-03-27T17:59:04Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"1b2821dacd549ffcdb8bb5ef3ce1a189b3049428dec8094118c5782bc1faa3f2","abstract_canon_sha256":"030c8c81143653284ec3987ecc36ba433bafefbcc26cdce4dbb69bf2a50ae1f7"},"schema_version":"1.0"},"canonical_sha256":"e0ee3393a767593f05b4295671c3f7ca5d4b2e57b7aa8c9cecb3d4cdd42fdf2a","source":{"kind":"arxiv","id":"2403.18814","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2403.18814","created_at":"2026-05-17T23:38:14Z"},{"alias_kind":"arxiv_version","alias_value":"2403.18814v1","created_at":"2026-05-17T23:38:14Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.18814","created_at":"2026-05-17T23:38:14Z"},{"alias_kind":"pith_short_12","alias_value":"4DXDHE5HM5MT","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"4DXDHE5HM5MT6BNU","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"4DXDHE5H","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:4DXDHE5HM5MT6BNUFFLHDQ7XZJ","target":"record","payload":{"canonical_record":{"source":{"id":"2403.18814","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-03-27T17:59:04Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"1b2821dacd549ffcdb8bb5ef3ce1a189b3049428dec8094118c5782bc1faa3f2","abstract_canon_sha256":"030c8c81143653284ec3987ecc36ba433bafefbcc26cdce4dbb69bf2a50ae1f7"},"schema_version":"1.0"},"canonical_sha256":"e0ee3393a767593f05b4295671c3f7ca5d4b2e57b7aa8c9cecb3d4cdd42fdf2a","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:38:14.687544Z","signature_b64":"bWyFaH6wDtqeefW78jGfFOysTjbmagGcj3hFZ6bPMKvxEb3tmHRwC7qP2U3VN2zNIpz+HSEQJV82GnR7HkoCBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e0ee3393a767593f05b4295671c3f7ca5d4b2e57b7aa8c9cecb3d4cdd42fdf2a","last_reissued_at":"2026-05-17T23:38:14.686781Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:38:14.686781Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2403.18814","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:38:14Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"okWTYm1XDBCVMWIxJhb9Bjxa3Ys7dr5tCtnT8pKjZWVJ24daCrl3T3nNKqHzPXRyJTzjO5wpT2O0G8Na8bHLDQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-05T10:11:11.608299Z"},"content_sha256":"7c6ed8ac9168d1c127e85a7842715b3eccb5affbd0f1ffc63af4b379833d5c68","schema_version":"1.0","event_id":"sha256:7c6ed8ac9168d1c127e85a7842715b3eccb5affbd0f1ffc63af4b379833d5c68"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:4DXDHE5HM5MT6BNUFFLHDQ7XZJ","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Mini-Gemini: Mining the Potential of Multi-modality Vision Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Mini-Gemini narrows the gap with top vision-language models by adding high-resolution image handling, better data, and self-guided generation without raising token counts.","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.CV","authors_text":"Chengyao Wang, Jiaya Jia, Ruihang Chu, Shaoteng Liu, Yanwei Li, Yixin Chen, Yuechen Zhang, Zhisheng Zhong","submitted_at":"2024-03-27T17:59:04Z","abstract_excerpt":"In this work, we introduce Mini-Gemini, a simple and effective framework enhancing multi-modality Vision Language Models (VLMs). Despite the advancements in VLMs facilitating basic visual dialog and reasoning, a performance gap persists compared to advanced models like GPT-4 and Gemini. We try to narrow the gap by mining the potential of VLMs for better performance and any-to-any workflow from three aspects, i.e., high-resolution visual tokens, high-quality data, and VLM-guided generation. To enhance visual tokens, we propose to utilize an additional visual encoder for high-resolution refineme"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Mini-Gemini is demonstrated to achieve leading performance in several zero-shot benchmarks and even surpasses the developed private models.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The performance improvements stem primarily from the three proposed components (high-resolution encoder, high-quality dataset, VLM-guided generation) rather than from undisclosed training details, hyperparameter tuning, or benchmark selection.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Mini-Gemini enhances VLMs via high-resolution visual refinement, curated reasoning data, and self-guided generation to reach leading zero-shot benchmark results across 2B-34B LLMs.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Mini-Gemini narrows the gap with top vision-language models by adding high-resolution image handling, better data, and self-guided generation without raising token counts.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"6b161878e9c5f0c28cd58566dec4beb744eb1b34395d6ec0ebdb871221e4ef52"},"source":{"id":"2403.18814","kind":"arxiv","version":1},"verdict":{"id":"35015c94-8dfe-49c3-997b-58dd097ddca1","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-17T07:39:59.742394Z","strongest_claim":"Mini-Gemini is demonstrated to achieve leading performance in several zero-shot benchmarks and even surpasses the developed private models.","one_line_summary":"Mini-Gemini enhances VLMs via high-resolution visual refinement, curated reasoning data, and self-guided generation to reach leading zero-shot benchmark results across 2B-34B LLMs.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The performance improvements stem primarily from the three proposed components (high-resolution encoder, high-quality dataset, VLM-guided generation) rather than from undisclosed training details, hyperparameter tuning, or benchmark selection.","pith_extraction_headline":"Mini-Gemini narrows the gap with top vision-language models by adding high-resolution image handling, better data, and self-guided generation without raising token counts."},"references":{"count":71,"sample":[{"doi":"","year":2023,"title":"OpenAI. Chatgpt. https://openai.com/blog/chatgpt/, 2023. 2","work_id":"88fc09ac-d3c3-42ba-bf70-f281d07aa22d","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"OPT: Open Pre-trained Transformer Language Models","work_id":"d7ff3b21-1fff-4cf4-952a-4714e3ef2307","ref_index":2,"cited_arxiv_id":"2205.01068","is_internal_anchor":true},{"doi":"","year":null,"title":"LLaMA: Open and Efficient Foundation Language Models","work_id":"c018fc23-6f3f-4035-9d02-28a2173b2b9d","ref_index":3,"cited_arxiv_id":"2302.13971","is_internal_anchor":true},{"doi":"","year":2023,"title":"GPT-4 Technical Report","work_id":"b928e041-6991-4c08-8c81-0359e4097c7b","ref_index":4,"cited_arxiv_id":"2303.08774","is_internal_anchor":true},{"doi":"","year":2023,"title":"Gemini: A Family of Highly Capable Multimodal Models","work_id":"83f7c85b-3f11-450f-ac0c-64d9745220b2","ref_index":5,"cited_arxiv_id":"2312.11805","is_internal_anchor":true}],"resolved_work":71,"snapshot_sha256":"ab6a103523b91a8f1f4e96612058cb9cfdadab5f14a5f758a075f0a072d84fd0","internal_anchors":27},"formal_canon":{"evidence_count":2,"snapshot_sha256":"2d4fab82c02fecd409097ab1bcc97a75aaa292c3d1101459d61c63a43aff2a67"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"35015c94-8dfe-49c3-997b-58dd097ddca1"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:38:14Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"PHt5C5mxb5GVgCRF/aHU+2YLJ5CuV8jN2ew+lgtPvPlzib7jrfvfOSHsdFHnymUrzKMk2rrDiS+qS2qZaIqADA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-05T10:11:11.608884Z"},"content_sha256":"79020ee7201756f9bd72c5845fd44d5674dd7115b0d92bc3f94317ac16446ed9","schema_version":"1.0","event_id":"sha256:79020ee7201756f9bd72c5845fd44d5674dd7115b0d92bc3f94317ac16446ed9"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/4DXDHE5HM5MT6BNUFFLHDQ7XZJ/bundle.json","state_url":"https://pith.science/pith/4DXDHE5HM5MT6BNUFFLHDQ7XZJ/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/4DXDHE5HM5MT6BNUFFLHDQ7XZJ/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-06-05T10:11:11Z","links":{"resolver":"https://pith.science/pith/4DXDHE5HM5MT6BNUFFLHDQ7XZJ","bundle":"https://pith.science/pith/4DXDHE5HM5MT6BNUFFLHDQ7XZJ/bundle.json","state":"https://pith.science/pith/4DXDHE5HM5MT6BNUFFLHDQ7XZJ/state.json","well_known_bundle":"https://pith.science/.well-known/pith/4DXDHE5HM5MT6BNUFFLHDQ7XZJ/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:4DXDHE5HM5MT6BNUFFLHDQ7XZJ","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":"030c8c81143653284ec3987ecc36ba433bafefbcc26cdce4dbb69bf2a50ae1f7","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-03-27T17:59:04Z","title_canon_sha256":"1b2821dacd549ffcdb8bb5ef3ce1a189b3049428dec8094118c5782bc1faa3f2"},"schema_version":"1.0","source":{"id":"2403.18814","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2403.18814","created_at":"2026-05-17T23:38:14Z"},{"alias_kind":"arxiv_version","alias_value":"2403.18814v1","created_at":"2026-05-17T23:38:14Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.18814","created_at":"2026-05-17T23:38:14Z"},{"alias_kind":"pith_short_12","alias_value":"4DXDHE5HM5MT","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"4DXDHE5HM5MT6BNU","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"4DXDHE5H","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:79020ee7201756f9bd72c5845fd44d5674dd7115b0d92bc3f94317ac16446ed9","target":"graph","created_at":"2026-05-17T23:38:14Z","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":"Mini-Gemini is demonstrated to achieve leading performance in several zero-shot benchmarks and even surpasses the developed private models."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The performance improvements stem primarily from the three proposed components (high-resolution encoder, high-quality dataset, VLM-guided generation) rather than from undisclosed training details, hyperparameter tuning, or benchmark selection."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Mini-Gemini enhances VLMs via high-resolution visual refinement, curated reasoning data, and self-guided generation to reach leading zero-shot benchmark results across 2B-34B LLMs."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Mini-Gemini narrows the gap with top vision-language models by adding high-resolution image handling, better data, and self-guided generation without raising token counts."}],"snapshot_sha256":"6b161878e9c5f0c28cd58566dec4beb744eb1b34395d6ec0ebdb871221e4ef52"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"2d4fab82c02fecd409097ab1bcc97a75aaa292c3d1101459d61c63a43aff2a67"},"paper":{"abstract_excerpt":"In this work, we introduce Mini-Gemini, a simple and effective framework enhancing multi-modality Vision Language Models (VLMs). Despite the advancements in VLMs facilitating basic visual dialog and reasoning, a performance gap persists compared to advanced models like GPT-4 and Gemini. We try to narrow the gap by mining the potential of VLMs for better performance and any-to-any workflow from three aspects, i.e., high-resolution visual tokens, high-quality data, and VLM-guided generation. To enhance visual tokens, we propose to utilize an additional visual encoder for high-resolution refineme","authors_text":"Chengyao Wang, Jiaya Jia, Ruihang Chu, Shaoteng Liu, Yanwei Li, Yixin Chen, Yuechen Zhang, Zhisheng Zhong","cross_cats":["cs.AI","cs.CL"],"headline":"Mini-Gemini narrows the gap with top vision-language models by adding high-resolution image handling, better data, and self-guided generation without raising token counts.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-03-27T17:59:04Z","title":"Mini-Gemini: Mining the Potential of Multi-modality Vision Language Models"},"references":{"count":71,"internal_anchors":27,"resolved_work":71,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"OpenAI. Chatgpt. https://openai.com/blog/chatgpt/, 2023. 2","work_id":"88fc09ac-d3c3-42ba-bf70-f281d07aa22d","year":2023},{"cited_arxiv_id":"2205.01068","doi":"","is_internal_anchor":true,"ref_index":2,"title":"OPT: Open Pre-trained Transformer Language Models","work_id":"d7ff3b21-1fff-4cf4-952a-4714e3ef2307","year":2022},{"cited_arxiv_id":"2302.13971","doi":"","is_internal_anchor":true,"ref_index":3,"title":"LLaMA: Open and Efficient Foundation Language Models","work_id":"c018fc23-6f3f-4035-9d02-28a2173b2b9d","year":null},{"cited_arxiv_id":"2303.08774","doi":"","is_internal_anchor":true,"ref_index":4,"title":"GPT-4 Technical Report","work_id":"b928e041-6991-4c08-8c81-0359e4097c7b","year":2023},{"cited_arxiv_id":"2312.11805","doi":"","is_internal_anchor":true,"ref_index":5,"title":"Gemini: A Family of Highly Capable Multimodal Models","work_id":"83f7c85b-3f11-450f-ac0c-64d9745220b2","year":2023}],"snapshot_sha256":"ab6a103523b91a8f1f4e96612058cb9cfdadab5f14a5f758a075f0a072d84fd0"},"source":{"id":"2403.18814","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-17T07:39:59.742394Z","id":"35015c94-8dfe-49c3-997b-58dd097ddca1","model_set":{"reader":"grok-4.3"},"one_line_summary":"Mini-Gemini enhances VLMs via high-resolution visual refinement, curated reasoning data, and self-guided generation to reach leading zero-shot benchmark results across 2B-34B LLMs.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Mini-Gemini narrows the gap with top vision-language models by adding high-resolution image handling, better data, and self-guided generation without raising token counts.","strongest_claim":"Mini-Gemini is demonstrated to achieve leading performance in several zero-shot benchmarks and even surpasses the developed private models.","weakest_assumption":"The performance improvements stem primarily from the three proposed components (high-resolution encoder, high-quality dataset, VLM-guided generation) rather than from undisclosed training details, hyperparameter tuning, or benchmark selection."}},"verdict_id":"35015c94-8dfe-49c3-997b-58dd097ddca1"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:7c6ed8ac9168d1c127e85a7842715b3eccb5affbd0f1ffc63af4b379833d5c68","target":"record","created_at":"2026-05-17T23:38:14Z","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":"030c8c81143653284ec3987ecc36ba433bafefbcc26cdce4dbb69bf2a50ae1f7","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-03-27T17:59:04Z","title_canon_sha256":"1b2821dacd549ffcdb8bb5ef3ce1a189b3049428dec8094118c5782bc1faa3f2"},"schema_version":"1.0","source":{"id":"2403.18814","kind":"arxiv","version":1}},"canonical_sha256":"e0ee3393a767593f05b4295671c3f7ca5d4b2e57b7aa8c9cecb3d4cdd42fdf2a","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"e0ee3393a767593f05b4295671c3f7ca5d4b2e57b7aa8c9cecb3d4cdd42fdf2a","first_computed_at":"2026-05-17T23:38:14.686781Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:38:14.686781Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"bWyFaH6wDtqeefW78jGfFOysTjbmagGcj3hFZ6bPMKvxEb3tmHRwC7qP2U3VN2zNIpz+HSEQJV82GnR7HkoCBQ==","signature_status":"signed_v1","signed_at":"2026-05-17T23:38:14.687544Z","signed_message":"canonical_sha256_bytes"},"source_id":"2403.18814","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:7c6ed8ac9168d1c127e85a7842715b3eccb5affbd0f1ffc63af4b379833d5c68","sha256:79020ee7201756f9bd72c5845fd44d5674dd7115b0d92bc3f94317ac16446ed9"],"state_sha256":"b8daf4361e060975d23ac705ed319fdb01b00160ae7f520cdc6647cff3cc422c"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"85kRUWcyZbHksp33eNyiBXwXsvIEgBQSd5v94JpgDwPg2GOvMCb2wb7DI+la77kPD8bYixyzvfyg7P3qgprJBw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-05T10:11:11.611660Z","bundle_sha256":"e194586139ae5b07f3656d166b570884c99f2474c8d419a564a343ebce439962"}}