{"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/transparency-by-design-closing-the-gap","title":"Transparency by Design: Closing the Gap Between Performance and Interpretability in Visual Reasoning","arxiv_id":"1803.05268","date":"2018-03-14","proceeding":"CVPR 2018 6","authors":["David Mascharka","Philip Tran","Ryan Soklaski","Arjun Majumdar"],"abstract":"Visual question answering requires high-order reasoning about an image, which\nis a fundamental capability needed by machine systems to follow complex\ndirectives. Recently, modular networks have been shown to be an effective\nframework for performing visual reasoning tasks. While modular networks were\ninitially designed with a degree of model transparency, their performance on\ncomplex visual reasoning benchmarks was lacking. Current state-of-the-art\napproaches do not provide an effective mechanism for understanding the\nreasoning process. In this paper, we close the performance gap between\ninterpretable models and state-of-the-art visual reasoning methods. We propose\na set of visual-reasoning primitives which, when composed, manifest as a model\ncapable of performing complex reasoning tasks in an explicitly-interpretable\nmanner. The fidelity and interpretability of the primitives' outputs enable an\nunparalleled ability to diagnose the strengths and weaknesses of the resulting\nmodel. Critically, we show that these primitives are highly performant,\nachieving state-of-the-art accuracy of 99.1% on the CLEVR dataset. We also show\nthat our model is able to effectively learn generalized representations when\nprovided a small amount of data containing novel object attributes. Using the\nCoGenT generalization task, we show more than a 20 percentage point improvement\nover the current state of the art.","url_abs":"http://arxiv.org/abs/1803.05268v2","url_pdf":"http://arxiv.org/pdf/1803.05268v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"transparency-by-design-closing-the-gap","repo_url":"https://github.com/davidmascharka/tbd-nets","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"question-answering","task_name":"Question Answering"},{"task_slug":"visual-question-answering-1","task_name":"Visual Question Answering"},{"task_slug":"visual-question-answering","task_name":"Visual Question Answering (VQA)"},{"task_slug":"visual-question-answering-vqa-split-a","task_name":"Visual Question Answering (VQA) Split A"},{"task_slug":"visual-question-answering-vqa-split-b","task_name":"Visual Question Answering (VQA) Split B"},{"task_slug":"visual-reasoning","task_name":"Visual Reasoning"}],"methods":[{"method_slug":"interpretability","method_name":"Interpretability"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/visual-question-answering-on-clevr","task":"Visual Question Answering (VQA)","dataset":"CLEVR","model":"TbD + reg + hres","rank_in_archive_order":5,"of":15,"metrics":{"Accuracy":"99.1"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1803.05268","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"1803.05268"}},"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":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/davidmascharka/tbd-nets","reach":{"status":"ok","spdx":"MIT"}}],"summary":{"unverified":8},"by_repo_kind":{"official":{"samples":8,"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":"30740fa63c238aeb","entry":"clevr_collate","repo":"davidmascharka/tbd-nets","repo_kind":"official","path":"utils/clevr.py","file_url":"https://github.com/davidmascharka/tbd-nets/blob/HEAD/utils/clevr.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":"30740fa63c238aeb"}},{"code_sha256_prefix":"8fb8b8481eaa4c9e","entry":"invert_dict","repo":"davidmascharka/tbd-nets","repo_kind":"official","path":"utils/clevr.py","file_url":"https://github.com/davidmascharka/tbd-nets/blob/HEAD/utils/clevr.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":"8fb8b8481eaa4c9e"}},{"code_sha256_prefix":"6efb4f0f662deed0","entry":"load_feature_extractor","repo":"davidmascharka/tbd-nets","repo_kind":"official","path":"utils/extract_features.py","file_url":"https://github.com/davidmascharka/tbd-nets/blob/HEAD/utils/extract_features.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":"6efb4f0f662deed0"}},{"code_sha256_prefix":"146d063ba938e454","entry":"load_program_generator","repo":"davidmascharka/tbd-nets","repo_kind":"official","path":"utils/generate_programs.py","file_url":"https://github.com/davidmascharka/tbd-nets/blob/HEAD/utils/generate_programs.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":"146d063ba938e454"}},{"code_sha256_prefix":"b9226c8e3adda1e5","entry":"load_tbd_net","repo":"davidmascharka/tbd-nets","repo_kind":"official","path":"tbd/module_net.py","file_url":"https://github.com/davidmascharka/tbd-nets/blob/HEAD/tbd/module_net.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":"b9226c8e3adda1e5"}},{"code_sha256_prefix":"ba3c072b123cf122","entry":"load_vocab","repo":"davidmascharka/tbd-nets","repo_kind":"official","path":"utils/clevr.py","file_url":"https://github.com/davidmascharka/tbd-nets/blob/HEAD/utils/clevr.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":"ba3c072b123cf122"}},{"code_sha256_prefix":"8fc4042c8274c184","entry":"logical_not","repo":"davidmascharka/tbd-nets","repo_kind":"official","path":"utils/generate_programs.py","file_url":"https://github.com/davidmascharka/tbd-nets/blob/HEAD/utils/generate_programs.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":"8fc4042c8274c184"}},{"code_sha256_prefix":"454f9b451952228a","entry":"logical_or","repo":"davidmascharka/tbd-nets","repo_kind":"official","path":"utils/generate_programs.py","file_url":"https://github.com/davidmascharka/tbd-nets/blob/HEAD/utils/generate_programs.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":"454f9b451952228a"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}