{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/arxiv-2605-00273","title":"When Do Diffusion Models learn to Generate Multiple Objects?","arxiv_id":"2605.00273","date":"2026-04-30","proceeding":"ICML","authors":["Yujin Jeong","Arnas Uselis","Iro Laina","Seong Joon Oh","Anna Rohrbach"],"abstract":"Text-to-image diffusion models achieve impressive visual fidelity, yet they remain unreliable in multi-object generation. Despite extensive empirical evidence of these failures, the underlying causes remain unclear. We begin by asking how much of this limitation arises from the data itself. To disentangle data effects, we consider two regimes across different dataset sizes: (1) concept generalization, where each individual concept is observed during training under potentially imbalanced data distributions, and (2) compositional generalization, where specific combinations of concepts are systematically held out. To study these regimes, we introduce mosaic (Multi-Object Spatial relations, AttrIbution, Counting), a controlled framework for dataset generation. By training diffusion models on mosaic, we find that scene complexity plays a dominant role rather than concept imbalance, and that counting is uniquely difficult to learn in low-data regimes. Moreover, compositional generalization collapses as more concept combinations are held out during training. These findings highlight fundamental limitations of diffusion models and motivate stronger inductive biases and data design for robust multi-object compositional generation.","url_abs":"https://arxiv.org/abs/2605.00273","url_pdf":"https://arxiv.org/pdf/2605.00273","source":{"archive":null,"snapshot":"2025-07-28","note":"not in the Papers with Code archive (frozen at the snapshot)","row_kind":"graph","title_abstract_authors_date":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)"},"code_links":[],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2605.00273","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2605.00273"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"mentioned_in_github":null,"is_official":null,"provenance":"deterministic:regex_extraction","mentioned_in_paper":null,"url":"https://github.com/eugene6923/MOSAIC","reach":null}],"summary":{"ran":11,"ran_draft_wrong":1,"ran_fixture":1,"unverified":4},"by_repo_kind":{"found_in_text":{"samples":17,"ran":13,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":17,"samples":[{"code_sha256_prefix":"119cf41aa9e4217f","entry":"accuracy_metric","repo":"eugene6923/MOSAIC","repo_kind":"found_in_text","path":"mosaic/comfort_utils/accuracy_metric.py","file_url":"https://github.com/eugene6923/MOSAIC/blob/HEAD/mosaic/comfort_utils/accuracy_metric.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"119cf41aa9e4217f"}},{"code_sha256_prefix":"8cf510a10c5f26c3","entry":"build_transform","repo":"eugene6923/MOSAIC","repo_kind":"found_in_text","path":"mosaic/comfort_utils/model_utils/intern_vl.py","file_url":"https://github.com/eugene6923/MOSAIC/blob/HEAD/mosaic/comfort_utils/model_utils/intern_vl.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"8cf510a10c5f26c3"}},{"code_sha256_prefix":"e8579f20cd680a34","entry":"calculate_tokens_mean_attn","repo":"eugene6923/MOSAIC","repo_kind":"found_in_text","path":"mosaic/comfort_utils/helper.py","file_url":"https://github.com/eugene6923/MOSAIC/blob/HEAD/mosaic/comfort_utils/helper.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"e8579f20cd680a34"}},{"code_sha256_prefix":"a721d26b902941ed","entry":"clean_latex_dataframe","repo":"eugene6923/MOSAIC","repo_kind":"found_in_text","path":"mosaic/comfort_utils/convert_perspective_taking_results.py","file_url":"https://github.com/eugene6923/MOSAIC/blob/HEAD/mosaic/comfort_utils/convert_perspective_taking_results.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"a721d26b902941ed"}},{"code_sha256_prefix":"d0aec2c5a6c35427","entry":"convert_to_dict","repo":"eugene6923/MOSAIC","repo_kind":"found_in_text","path":"mosaic/comfort_utils/convert_general_metrics_results.py","file_url":"https://github.com/eugene6923/MOSAIC/blob/HEAD/mosaic/comfort_utils/convert_general_metrics_results.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"d0aec2c5a6c35427"}},{"code_sha256_prefix":"d610b5eabe0c9db1","entry":"dynamic_preprocess","repo":"eugene6923/MOSAIC","repo_kind":"found_in_text","path":"mosaic/comfort_utils/model_utils/intern_vl.py","file_url":"https://github.com/eugene6923/MOSAIC/blob/HEAD/mosaic/comfort_utils/model_utils/intern_vl.py","link_basis":"plan_row","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"well_formed","behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"d610b5eabe0c9db1"}},{"code_sha256_prefix":"d3f911d03b9b9162","entry":"extract_table_data","repo":"eugene6923/MOSAIC","repo_kind":"found_in_text","path":"mosaic/comfort_utils/convert_general_metrics_results.py","file_url":"https://github.com/eugene6923/MOSAIC/blob/HEAD/mosaic/comfort_utils/convert_general_metrics_results.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"d3f911d03b9b9162"}},{"code_sha256_prefix":"bd77f5f8067f18e9","entry":"find_closest_aspect_ratio","repo":"eugene6923/MOSAIC","repo_kind":"found_in_text","path":"mosaic/comfort_utils/model_utils/intern_vl.py","file_url":"https://github.com/eugene6923/MOSAIC/blob/HEAD/mosaic/comfort_utils/model_utils/intern_vl.py","link_basis":"harvester_set","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":"well_formed","behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"bd77f5f8067f18e9"}},{"code_sha256_prefix":"69b3b0492d912e11","entry":"find_index_of_image_tokens","repo":"eugene6923/MOSAIC","repo_kind":"found_in_text","path":"mosaic/comfort_utils/helper.py","file_url":"https://github.com/eugene6923/MOSAIC/blob/HEAD/mosaic/comfort_utils/helper.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"69b3b0492d912e11"}},{"code_sha256_prefix":"55c9cd0497a84208","entry":"grounding_enc_processor","repo":"eugene6923/MOSAIC","repo_kind":"found_in_text","path":"mosaic/comfort_utils/model_utils/glamm_wrapper.py","file_url":"https://github.com/eugene6923/MOSAIC/blob/HEAD/mosaic/comfort_utils/model_utils/glamm_wrapper.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"55c9cd0497a84208"}},{"code_sha256_prefix":"50dff60708886cd3","entry":"latex_to_dataframe","repo":"eugene6923/MOSAIC","repo_kind":"found_in_text","path":"mosaic/comfort_utils/convert_perspective_taking_results.py","file_url":"https://github.com/eugene6923/MOSAIC/blob/HEAD/mosaic/comfort_utils/convert_perspective_taking_results.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"50dff60708886cd3"}},{"code_sha256_prefix":"c8d3c86f905b4d9a","entry":"sort_dataframe_by_model","repo":"eugene6923/MOSAIC","repo_kind":"found_in_text","path":"mosaic/comfort_utils/convert_perspective_taking_results.py","file_url":"https://github.com/eugene6923/MOSAIC/blob/HEAD/mosaic/comfort_utils/convert_perspective_taking_results.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"c8d3c86f905b4d9a"}},{"code_sha256_prefix":"99bc5e82093e6f29","entry":"sub_string_tokens_in_resp_token","repo":"eugene6923/MOSAIC","repo_kind":"found_in_text","path":"mosaic/comfort_utils/model_utils/gpt_wrapper.py","file_url":"https://github.com/eugene6923/MOSAIC/blob/HEAD/mosaic/comfort_utils/model_utils/gpt_wrapper.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"99bc5e82093e6f29"}},{"code_sha256_prefix":"2846873087c38293","entry":"accuracy_metric_object_frame","repo":"eugene6923/MOSAIC","repo_kind":"found_in_text","path":"mosaic/comfort_utils/accuracy_metric.py","file_url":"https://github.com/eugene6923/MOSAIC/blob/HEAD/mosaic/comfort_utils/accuracy_metric.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"2846873087c38293"}},{"code_sha256_prefix":"0497e51919a7a2a6","entry":"accuracy_metric_old","repo":"eugene6923/MOSAIC","repo_kind":"found_in_text","path":"mosaic/comfort_utils/accuracy_metric.py","file_url":"https://github.com/eugene6923/MOSAIC/blob/HEAD/mosaic/comfort_utils/accuracy_metric.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"0497e51919a7a2a6"}},{"code_sha256_prefix":"c68d30db2cd63b2a","entry":"convention_metric","repo":"eugene6923/MOSAIC","repo_kind":"found_in_text","path":"mosaic/comfort_utils/convention_metric.py","file_url":"https://github.com/eugene6923/MOSAIC/blob/HEAD/mosaic/comfort_utils/convention_metric.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"c68d30db2cd63b2a"}},{"code_sha256_prefix":"c36a70c23b159d96","entry":"language_ambiguity_metric","repo":"eugene6923/MOSAIC","repo_kind":"found_in_text","path":"mosaic/comfort_utils/language_ambiguity_metric.py","file_url":"https://github.com/eugene6923/MOSAIC/blob/HEAD/mosaic/comfort_utils/language_ambiguity_metric.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"c36a70c23b159d96"}}]},"arxiv_metadata":{"licence":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)","fields":["title","abstract","authors","date"],"primary_category":"cs.CV","source":"arxiv_2026.jsonl"},"syntology_extracted_results":null}