{"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-2603-13421","title":"Generalization and Memorization in Rectified Flow","arxiv_id":"2603.13421","date":"2026-03-12","proceeding":null,"authors":["Mingxing Rao","Daniel Moyer"],"abstract":"Generative models based on the Flow Matching objective, particularly Rectified Flow, have emerged as a dominant paradigm for efficient, high-fidelity image synthesis. However, while existing research heavily prioritizes generation quality and architectural scaling, the underlying dynamics of how RF models memorize training data remain largely underexplored. In this paper, we systematically investigate the memorization behaviors of RF through the test statistics of Membership Inference Attacks (MIA). We progressively formulate three test statistics, culminating in a complexity-calibrated metric that successfully decouples intrinsic image spatial complexity from genuine memorization signals. This calibration yields a significant performance surge -- boosting attack AUC by up to 15% and the privacy-critical TPR@1%FPR metric by up to 45% -- establishing the first non-trivial MIA specifically tailored for RF. Leveraging these refined metrics, we uncover a distinct temporal pattern: under standard uniform temporal training, a model's susceptibility to MIA strictly peaks at the integration midpoint, a phenomenon we justify via the network's forced deviation from linear approximations. Finally, we demonstrate that substituting uniform timestep sampling with a Symmetric Exponential (U-shaped) distribution effectively minimizes exposure to vulnerable intermediate timesteps. Extensive evaluations across three datasets confirm that this temporal regularization suppresses memorization while preserving generative fidelity.","url_abs":"https://arxiv.org/abs/2603.13421","url_pdf":"https://arxiv.org/pdf/2603.13421","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":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=2603.13421","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2603.13421"}},"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/mx-ethan-rao/rf_gen_mem","reach":null}],"summary":{"unverified":16},"by_repo_kind":{"found_in_text":{"samples":16,"ran":0,"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":0,"samples":[{"code_sha256_prefix":"12ac890d58be85af","entry":"calculate_mia_auc","repo":"mx-ethan-rao/rf_gen_mem","repo_kind":"found_in_text","path":"src/cal_statistics_cifar10.py","file_url":"https://github.com/mx-ethan-rao/rf_gen_mem/blob/HEAD/src/cal_statistics_cifar10.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"12ac890d58be85af"}},{"code_sha256_prefix":"47417e74e8a39713","entry":"calculate_mia_auc","repo":"mx-ethan-rao/rf_gen_mem","repo_kind":"found_in_text","path":"src/cal_statistics_cifar10_dpsgd.py","file_url":"https://github.com/mx-ethan-rao/rf_gen_mem/blob/HEAD/src/cal_statistics_cifar10_dpsgd.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"47417e74e8a39713"}},{"code_sha256_prefix":"a52425ad738321d1","entry":"calculate_mia_auc","repo":"mx-ethan-rao/rf_gen_mem","repo_kind":"found_in_text","path":"src/cal_statistics_imnet.py","file_url":"https://github.com/mx-ethan-rao/rf_gen_mem/blob/HEAD/src/cal_statistics_imnet.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"a52425ad738321d1"}},{"code_sha256_prefix":"5001529b2bcda5e7","entry":"calculate_mia_auc","repo":"mx-ethan-rao/rf_gen_mem","repo_kind":"found_in_text","path":"src/cal_statistics_svhn.py","file_url":"https://github.com/mx-ethan-rao/rf_gen_mem/blob/HEAD/src/cal_statistics_svhn.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"5001529b2bcda5e7"}},{"code_sha256_prefix":"17709e6d0601c6b1","entry":"compute_mc_scores","repo":"mx-ethan-rao/rf_gen_mem","repo_kind":"found_in_text","path":"src/attack_latent_mscoco.py","file_url":"https://github.com/mx-ethan-rao/rf_gen_mem/blob/HEAD/src/attack_latent_mscoco.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"17709e6d0601c6b1"}},{"code_sha256_prefix":"f63c6ca58ea090b4","entry":"compute_tpr_at_fpr","repo":"mx-ethan-rao/rf_gen_mem","repo_kind":"found_in_text","path":"src/attack_cifar10.py","file_url":"https://github.com/mx-ethan-rao/rf_gen_mem/blob/HEAD/src/attack_cifar10.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"f63c6ca58ea090b4"}},{"code_sha256_prefix":"0aaaf0d6b8795657","entry":"compute_tpr_at_fpr","repo":"mx-ethan-rao/rf_gen_mem","repo_kind":"found_in_text","path":"src/attack_latent_mscoco.py","file_url":"https://github.com/mx-ethan-rao/rf_gen_mem/blob/HEAD/src/attack_latent_mscoco.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"0aaaf0d6b8795657"}},{"code_sha256_prefix":"92f7bd4d42444820","entry":"evaluate_loader","repo":"mx-ethan-rao/rf_gen_mem","repo_kind":"found_in_text","path":"src/cal_train_val_loss.py","file_url":"https://github.com/mx-ethan-rao/rf_gen_mem/blob/HEAD/src/cal_train_val_loss.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"92f7bd4d42444820"}},{"code_sha256_prefix":"ad3ead733ddd1366","entry":"get_complexity_scores","repo":"mx-ethan-rao/rf_gen_mem","repo_kind":"found_in_text","path":"src/cal_statistics_cifar10.py","file_url":"https://github.com/mx-ethan-rao/rf_gen_mem/blob/HEAD/src/cal_statistics_cifar10.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"ad3ead733ddd1366"}},{"code_sha256_prefix":"fc0545d8b5d98fb1","entry":"get_complexity_scores","repo":"mx-ethan-rao/rf_gen_mem","repo_kind":"found_in_text","path":"src/cal_statistics_cifar10_dpsgd.py","file_url":"https://github.com/mx-ethan-rao/rf_gen_mem/blob/HEAD/src/cal_statistics_cifar10_dpsgd.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"fc0545d8b5d98fb1"}},{"code_sha256_prefix":"cdbff466b2dfcb94","entry":"get_complexity_scores","repo":"mx-ethan-rao/rf_gen_mem","repo_kind":"found_in_text","path":"src/cal_statistics_imnet.py","file_url":"https://github.com/mx-ethan-rao/rf_gen_mem/blob/HEAD/src/cal_statistics_imnet.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"cdbff466b2dfcb94"}},{"code_sha256_prefix":"84590f01d4800f95","entry":"get_complexity_scores","repo":"mx-ethan-rao/rf_gen_mem","repo_kind":"found_in_text","path":"src/cal_statistics_svhn.py","file_url":"https://github.com/mx-ethan-rao/rf_gen_mem/blob/HEAD/src/cal_statistics_svhn.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"84590f01d4800f95"}},{"code_sha256_prefix":"73aa1c85f33c8b60","entry":"get_complexity_scores","repo":"mx-ethan-rao/rf_gen_mem","repo_kind":"found_in_text","path":"src/cal_train_val_loss.py","file_url":"https://github.com/mx-ethan-rao/rf_gen_mem/blob/HEAD/src/cal_train_val_loss.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"73aa1c85f33c8b60"}},{"code_sha256_prefix":"efdbbf7f5d6788a7","entry":"get_complexity_scores","repo":"mx-ethan-rao/rf_gen_mem","repo_kind":"found_in_text","path":"src/likelihood_attack_calibrated.py","file_url":"https://github.com/mx-ethan-rao/rf_gen_mem/blob/HEAD/src/likelihood_attack_calibrated.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"efdbbf7f5d6788a7"}},{"code_sha256_prefix":"4f9090b4d2c5949f","entry":"get_step","repo":"mx-ethan-rao/rf_gen_mem","repo_kind":"found_in_text","path":"src/cal_train_val_loss.py","file_url":"https://github.com/mx-ethan-rao/rf_gen_mem/blob/HEAD/src/cal_train_val_loss.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"4f9090b4d2c5949f"}},{"code_sha256_prefix":"48550d1d189bf685","entry":"preprocess_eval","repo":"mx-ethan-rao/rf_gen_mem","repo_kind":"found_in_text","path":"src/attack_latent_mscoco.py","file_url":"https://github.com/mx-ethan-rao/rf_gen_mem/blob/HEAD/src/attack_latent_mscoco.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"48550d1d189bf685"}}]},"arxiv_metadata":{"licence":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)","fields":["title","abstract","authors","date"],"primary_category":"cs.LG","source":"arxiv_2026.jsonl"},"syntology_extracted_results":null}