{"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":"/census/perturb","entry":"perturb","source":"Syntology differential census (groundwork/55, run v2_2026-09-22), per sample; not an archive number","census_date":"2026-09-22","battery_sha256":["b986f7e04d794a0d88ad4c5f32cf63ec3590b5deff0192150737bc5f1c0b4677"],"runner_sha256":["5a452d0e7c0da5b80771d1be2afe3572e253568cf5c5d59a0d08ebd663d00808"],"bucket_key":"positional (rank, kind, dtype) of each array argument; the argument name is not part of the key because the harness draws the shared array from (rank, kind) and casts it to the dtype, whatever the name","bucket_fields":["rank","kind","dtype"],"claim":"Implementations sharing this entry name were each run on one shared input fixed by the positional (rank, kind, dtype) of their array arguments (the bucket). A cluster is the set whose recorded output digest (sha256 of the output rounded to 6 decimals) is identical. Identical values to six decimals on the shared input are agreement on those inputs, not a statement about the whole domain and not a substitution claim.","n_implementations_compared":4,"n_papers":5,"n_buckets":2,"n_distinct_outputs":4,"n_class_bearing":0,"not_compared":{"not_run_on_shared_input":{"n":0,"by_error":{}},"no_array_argument_ran_on_own_fixture_arguments_only":{"n":1,"n_papers":1,"recorded_shared_digest_equals_own_fixture_digest":1},"output_not_digested_non_numeric":{"n":0,"by_type":{}}},"withdrawn_excluded":0,"code_page":"/code/perturb","buckets":[{"bucket":[[2,"float","float32"]],"n_implementations_compared":2,"n_papers":2,"n_not_digested":0,"not_digested_by_type":{},"clusters":[{"output_sha":"bf13ebefde930547","size":1,"n_papers":1,"shape":[4,8],"dtype":"float32","type":"Tensor","finite":true,"max_difference_among_recorded_values":null,"members_with_recorded_values":1,"torch_reference_conventions_with_this_digest":[],"values":[1.690044641494751,-1.6490123271942139,0.45099538564682007,-0.7342749238014221,-0.8179784417152405,-0.961270272731781,-0.6943621635437012,0.36136722564697266,0.5511034727096558,0.8328121900558472,-0.3622012436389923,-1.5375348329544067,-0.27655094861984253,2.4458744525909424,-0.4031062722206116,0.5258008241653442,-0.0688699409365654,-1.7924686670303345,0.48877647519111633,-0.326835036277771,0.03728972375392914,-1.3362919092178345,-0.571004331111908,1.168974757194519,-0.7205938696861267,0.37472471594810486,0.927638053894043,1.0358660221099854,-1.0868490934371948,1.4328514337539673,-1.2275919914245605,-0.6784768104553223],"values_recorded":32,"members":[{"code_sha256_prefix":"f4e22a0a3a8ec839","path":"pretraining.py","papers":["2309.15718"],"paper_pages":[{"arxiv_id":"2309.15718","page":"/paper/gpip-geometry-enhanced-pre-training-on"}],"arg_sig_recorded":[["positions",2,"float","float32"]],"scalar_args":{"mu":"0.0","sigma":"0.1"},"class_bearing":false}]},{"output_sha":"c47a54b9b221a62b","size":1,"n_papers":1,"shape":[4,8],"dtype":"float32","type":"ndarray","finite":true,"max_difference_among_recorded_values":null,"members_with_recorded_values":1,"torch_reference_conventions_with_this_digest":[],"values":[2.684654712677002,-1.3336976766586304,0.9654222130775452,0.4295595586299896,0.030929500237107277,-1.5191102027893066,-0.18771670758724213,0.4972105920314789,0.4672665297985077,1.164444923400879,-0.32517778873443604,-0.8412114381790161,0.09198377281427383,2.382946252822876,-0.29285237193107605,0.7173657417297363,0.8134349584579468,-1.7254546880722046,0.5886452794075012,-0.8332337141036987,-1.2990890741348267,-0.8539731502532959,-0.10465017706155777,0.6125916838645935,0.33926451206207275,-0.29390841722488403,0.9678570032119751,0.9239263534545898,-0.4593961238861084,2.008897304534912,-1.2447481155395508,-0.4050278663635254],"values_recorded":32,"members":[{"code_sha256_prefix":"2924888d72c2c274","path":"baselines/her/her.py","papers":["1906.05838"],"paper_pages":[{"arxiv_id":"1906.05838","page":"/paper/goal-conditioned-imitation-learning"}],"arg_sig_recorded":[["states",2,"float","float32"]],"scalar_args":{"scale":"0.5"},"class_bearing":false}]}]},{"bucket":[[4,"float","float32"]],"n_implementations_compared":2,"n_papers":3,"n_not_digested":0,"not_digested_by_type":{},"clusters":[{"output_sha":"d8dff36cd4ec80e1","size":1,"n_papers":2,"shape":[2,3,4,4],"dtype":"float32","type":"Tensor","finite":true,"max_difference_among_recorded_values":null,"members_with_recorded_values":1,"torch_reference_conventions_with_this_digest":[],"values":[0.7231047749519348,0.6530793309211731,1.4460861682891846,1.0985400676727295,1.1372957229614258,-0.19992724061012268,-0.2714869976043701,0.8319791555404663,0.3433533310890198,-0.13238386809825897,-1.1905570030212402,-0.012713957577943802,-0.709074079990387,-0.4362029433250427,-2.665717124938965,0.812962532043457,-0.45303258299827576,0.12553228437900543,0.5928415060043335,-1.463806390762329,-0.7250184416770935,1.1514394283294678,-0.3975509703159332,1.4335219860076904,0.22366678714752197,1.184694528579712,-2.0680742263793945,-0.16907231509685516,0.3742450773715973,0.00391392083838582,0.895301342010498,-0.4980470836162567,-0.4213716387748718,1.020633339881897,-0.4836711883544922,-0.4854681193828583,-0.6618722677230835,1.982889175415039,0.2125764787197113,-1.2040232419967651,-0.21204684674739838,-1.479430913925171,-0.7425822019577026,1.2173826694488525,0.022702597081661224,-0.16384494304656982,0.3325777053833008,-1.246457576751709,1.0319796800613403,0.47569993138313293,2.1016592979431152,-0.01877530850470066,1.654322624206543,-0.6244865655899048,-2.408447742462158,-0.21435445547103882,-1.3870023488998413,-2.00654673576355,-1.3687102794647217,-0.026417039334774017,-2.796109437942505,-0.7385192513465881,-0.6348524689674377,0.21286723017692566,0.8440517783164978,-1.0907533168792725,0.7723703980445862,2.0942370891571045,0.4846275746822357,0.022118685767054558,3.1044223308563232,0.0736190527677536,0.6060231924057007,0.28421318531036377,0.5158193111419678,-0.6696537733078003,1.0927388668060303,0.267061322927475,0.6621701121330261,0.2371496558189392,-2.2168080806732178,-0.08749820291996002,0.16540314257144928,-0.8387367129325867,0.7167379260063171,1.2303640842437744,-1.5213911533355713,1.0778614282608032,0.22837327420711517,0.4391850531101227,1.9725817441940308,-1.1686570644378662,0.49750959873199463,0.14785639941692352,-1.3677366971969604,-0.28564757108688354],"values_recorded":96,"members":[{"code_sha256_prefix":"79c0d47a0e121164","path":"attack.py","papers":["1903.08333","1803.04765"],"paper_pages":[{"arxiv_id":"1903.08333","page":"/paper/on-the-robustness-of-deep-k-nearest-neighbors"},{"arxiv_id":"1803.04765","page":"/paper/deep-k-nearest-neighbors-towards-confident"}],"arg_sig_recorded":[["data",4,"float","float32"]],"scalar_args":{},"class_bearing":false}]},{"output_sha":"d5a71bf50f274ad6","size":1,"n_papers":1,"shape":[2,3,4,4],"dtype":"float32","type":"Tensor","finite":true,"max_difference_among_recorded_values":null,"members_with_recorded_values":1,"torch_reference_conventions_with_this_digest":[],"values":[0.7231047749519348,0.6530793309211731,0.2078431397676468,0.5490196347236633,1.1372957229614258,-0.19992724061012268,-0.2714869976043701,0.8319791555404663,0.3433533310890198,-0.13238386809825897,-1.1905570030212402,-0.012713957577943802,0.49803921580314636,0.95686274766922,0.5137255191802979,0.812962532043457,-0.45303258299827576,0.12553228437900543,0.9098039269447327,0.30588236451148987,-0.7250184416770935,1.1514394283294678,-0.3975509703159332,1.4335219860076904,0.22366678714752197,1.184694528579712,-2.0680742263793945,0.4588235318660736,0.3333333432674408,0.00391392083838582,0.895301342010498,-0.4980470836162567,0.6627451181411743,0.6392157077789307,0.7529411911964417,0.37254902720451355,0.772549033164978,1.982889175415039,0.2125764787197113,-1.2040232419967651,-0.21204684674739838,0.1411764770746231,-0.7425822019577026,0.1882352977991104,0.364705890417099,0.5137255191802979,0.3325777053833008,-1.246457576751709,0.8039215803146362,0.47569993138313293,2.1016592979431152,-0.01877530850470066,1.654322624206543,-0.6244865655899048,-2.408447742462158,0.5411764979362488,-1.3870023488998413,0.16862745583057404,0.729411780834198,0.49803921580314636,-2.796109437942505,-0.7385192513465881,0.5098039507865906,0.21286723017692566,0.8440517783164978,-1.0907533168792725,0.7723703980445862,2.0942370891571045,0.4846275746822357,0.022118685767054558,0.32156863808631897,0.0736190527677536,0.6060231924057007,0.28421318531036377,0.5158193111419678,-0.6696537733078003,1.0927388668060303,0.054901961237192154,0.6621701121330261,0.22745098173618317,0.7568627595901489,-0.08749820291996002,0.16540314257144928,0.33725491166114807,0.7167379260063171,1.2303640842437744,-1.5213911533355713,0.007843137718737125,0.22837327420711517,0.3137255012989044,1.9725817441940308,-1.1686570644378662,0.49750959873199463,0.14785639941692352,0.07450980693101883,0.6823529601097107],"values_recorded":96,"members":[{"code_sha256_prefix":"356aef4a4b74b441","path":"train_pixel.py","papers":["2003.02977"],"paper_pages":[{"arxiv_id":"2003.02977","page":"/paper/likelihood-regret-an-out-of-distribution"}],"arg_sig_recorded":[["x",4,"float","float32"]],"scalar_args":{"mu":"0.3","device":"device(type='cpu')"},"class_bearing":false}]}]}]}