softmax
softmax: 53 implementations from 63 papers ran on one shared input (census 2026-09-22, battery b986f7e04d79); they produced 16 distinct outputs across 8 buckets, one shared input per bucket.
Identical values to six decimals (the recorded digest) on the shared input are agreement on those inputs, not equivalence. Implementations are compared only within one bucket, the 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; each bucket's shared input is fixed by that key, so members of one bucket saw bitwise-identical inputs under their own scalar arguments. A cluster is the set of members whose recorded output digest is identical. Nothing here says which computation a paper's method intended, and nothing reproduces a paper's results. Each member's behaviour fingerprint (its outputs on the shared battery) is not in this export and is not shown here; get_code(code_sha256="…") returns this sample's fingerprint on the shared battery, with the battery's shuffled null and lift; not shown on this site (how to connect); click a member's code_sha256 to copy that call.
Not compared, and not in the tables or the counts above:
- 1 more implementation of this name was examined but did not run on the shared input (ValueError 1).
- 1 more implementation from 1 paper took no array argument and ran only on its own fixture arguments; no shared input existed for it, so it is not compared (on 1 the recorded shared-output digest is the own-fixture digest).
- Every compared output was digested.
- The harness also ran torch's own computation of this name under 6 named conventions; those rows are not implementations, are not counted, and only label a cluster whose digest they share.
Bucket 1 of 8: arg 1: rank 2, kind float, dtype float32
25 implementations from 25 papers share this bucket (rank, kind, dtype of each array argument, positional; each member's recorded signature, argument name included, is shown under it); 4 distinct outputs, largest cluster first. Values are the first 8 of the recorded output, flattened.
| Cluster (same digest to six decimals) | Members | Shared output on this bucket's input |
|---|---|---|
| 10 implementations 10 papers 95d860eeb84c recorded values differ by up to 1.49e-08 |
caliblab/calibrators/conformal_mass_threshold_calibrator.py 08bc835660d4723f recorded logits:2/float/float32; axis=-1 mr_uplift/erupt.py 74138c11e6beb035 recorded x:2/float/float32 utils.py 785d346b27b67e09 recorded x:2/float/float32; axis=1 src/benchmark/softmax.py 7c0d7e9e7a7f89cc recorded x:2/float/float32
code/stac.py 7ceb8f93a02b4ee3 recorded Z:2/float/float32 src/score_utils.py 7ddc824e88e78020 recorded x:2/float/float32 PACE/src/model.py 8d6c2bedc92a8075 recorded x:2/float/float32 utils.py a16c332f67a73877 recorded logp:2/float/float32; axis=-1 utils.py db83f3a3d4f7c604 recorded X:2/float/float32 src/grounding_evaluator.py f89fc809b3a7824a recorded x:2/float/float32 |
[0.530059, 0.0188513, 0.14067, 0.043793, 0.0354283, 0.0311835, 0.0450422, 0.154973, …] shape [4, 8] · float32 · ndarray |
| 8 implementations 9 papers 09f6f45b4d58 recorded values differ by up to 5.96e-08 same digest as torch: dim0 |
one code sha held from 2 papers' repositories bert/evaluate.py cb5cbea0e8219ff5 recorded x:2/float/float32 dreamerv2/goal_picker.py 083546dc5ac62494 recorded X:2/float/float32; theta=1.0, axis=None train_model.py 51f9969762b91ea3 recorded X:2/float/float32; theta=1.0, axis=None transformer.py 54fe686797cba4a9 recorded ys:2/float/float32 mDLAM.py 81767458c6ff4e4d recorded x:2/float/float32 1_easyrun/RankC.py a49ec0b422e77ed3 recorded x:2/float/float32 al_neural_dialogue_train.py b2d8244fa9563626 recorded x:2/float/float32 test_demo.py cd98260d20abd01f recorded x:2/float/float32; temp=1.0 |
[0.654629, 0.047257, 0.251707, 0.120987, 0.167114, 0.0247171, 0.262414, 0.263353, …] shape [4, 8] · float32 · ndarray |
| 6 implementations 5 papers ab42e90e344c recorded values differ by up to 1.19e-07 same digest as torch: last_dim |
soft_moe/soft_moe.py 348d3e3d68c0c2dc recorded x:2/float/float32; dim=1 src/train_eval.py 62d3b888dc672b51 recorded tensor:2/float/float32 soft_moe/soft_moe.py 84195cc106a2908b recorded x:2/float/float32; dim=1 lib/utils.py 8825c7c596e27b6f recorded x:2/float/float32 models/dgrmil.py c42e5fa46bf385ab recorded x:2/float/float32; dim=1, onnx_trace=False src/models/programs.py e3599c195e20c0fc recorded X:2/float/float32; dim=-1, tau=1.0 |
[0.530059, 0.0188513, 0.14067, 0.043793, 0.0354283, 0.0311835, 0.0450422, 0.154973, …] shape [4, 8] · float32 · Tensor |
| 1 implementation 1 paper 410404bd78e6 |
scripts/04_evaluate.py 6a445568164614eb recorded a:2/float/float32 |
[3.42957] shape [] · float64 · float |
Bucket 2 of 8: arg 1: rank 1, kind float, dtype float64
14 implementations from 23 papers share this bucket (rank, kind, dtype of each array argument, positional; each member's recorded signature, argument name included, is shown under it); 4 distinct outputs, largest cluster first. Values are the first 8 of the recorded output, flattened.
| Cluster (same digest to six decimals) | Members | Shared output on this bucket's input |
|---|---|---|
| 11 implementations 19 papers aa13e79c0065 recorded values differ by up to 1.09e-13 |
smoe/entrypoint/eval/eval_mmlu_moe_0.py 19f01570b0a18e2b recorded x:1/float/float64
random_mutation_search.py 7f9b15275ca5759c recorded x:1/float/float64
word2vec.py c0cf1196424d4f71 recorded x:1/float/float64
query_openai.py 9c4be464cbe63fba recorded x:1/float/float64
src/vlmeval/calculate_scores/map_predictions.py 08883768489fd226 recorded x:1/float/float64 search.py 380f691b7059126a recorded probs:1/float/float64 src/compute_results.py 4471357a1f78d539 recorded x:1/float/float64 mmd_dilated/mmd_dilated.py 50d5258bd0761c1b recorded x:1/float/float64 OpenAttack/attackers/genetic.py 54948e052031386e recorded inputs:1/float/float64
src/cp_methods.py 885d6e6994f5e4bd recorded x:1/float/float64 toy/ARSM_Univariate.py ad039f77ef1e89fc recorded phi:1/float/float64 |
[0.0545667, 0.157661, 0.330605, 0.062873, 0.112963, 0.142074, 0.0264399, 0.112818] shape [8] · float64 · ndarray |
| 1 implementation 2 papers 231a35f00878 |
one code sha held from 2 papers' repositories MAB/train_mab_mo_multiple.py 69e582d84850bb60 recorded data:1/float/float64; tau=1.2 |
[0.064659, 0.15654, 0.290144, 0.0727628, 0.118568, 0.143532, 0.0353513, 0.118442] shape [8] · float64 · ndarray |
| 1 implementation 1 paper 0c77544de6fb |
rl_algorithm/common/option_model.py 3be33e47b8058815 recorded a:1/float/float64; ww=0.5 |
[0.0875577, 0.14883, 0.215519, 0.0939858, 0.125979, 0.141282, 0.060948, 0.125898] shape [8] · float64 · ndarray |
| 1 implementation 1 paper 92bf9c23a889 |
discrete_usher/clean_q_implementation.py 4a6f0cbbf7e6a8d9 recorded arr:1/float/float64; temp=0.5 |
[0.0989503, 0.142928, 0.184744, 0.103931, 0.127332, 0.137863, 0.076977, 0.127275] shape [8] · float64 · ndarray |
Bucket 3 of 8: arg 1: rank 2, kind float, dtype float64
9 implementations from 12 papers share this bucket (rank, kind, dtype of each array argument, positional; each member's recorded signature, argument name included, is shown under it); 3 distinct outputs, largest cluster first. Values are the first 8 of the recorded output, flattened.
| Cluster (same digest to six decimals) | Members | Shared output on this bucket's input |
|---|---|---|
| 6 implementations 7 papers ab42e90e344c recorded values differ by up to 3.06e-11 |
one code sha held from 2 papers' repositories metrics.py a8ae5c6e4eb58226 recorded X:2/float/float64; copy=True yolo_detection.py 475ce08eea0c1d56 recorded val:2/float/float64; axis=-1 code/debias_utils.py 64e12b411d5ec4ca recorded x:2/float/float64 evaluation.py 652f07c5ff410713 recorded X:2/float/float64; axis=1 a2/word2vec.py 9374cbb282b14e3e recorded x:2/float/float64 logistic.py aa0acf499f826391 recorded x:2/float/float64 |
[0.530059, 0.0188513, 0.14067, 0.043793, 0.0354283, 0.0311835, 0.0450422, 0.154973, …] shape [4, 8] · float64 · ndarray |
| 2 implementations 3 papers 8f915a52849e recorded values identical |
COMMON/src/lap_solvers/ILP.py 98bba89c2a19a3bb recorded x:2/float/float64; axis=None generate_data.py d8e37876ec1be74c recorded x:2/float/float64; temp=1.0, axis=None |
[0.121485, 0.00432056, 0.0322402, 0.010037, 0.00811985, 0.00714698, 0.0103233, 0.0355184, …] shape [4, 8] · float64 · ndarray |
| 1 implementation 2 papers 09f6f45b4d58 |
image_classification/assistant.py 91fe7a23c1c052b7 recorded x:2/float/float64 |
[0.654629, 0.047257, 0.251707, 0.120987, 0.167114, 0.0247171, 0.262414, 0.263353, …] shape [4, 8] · float64 · ndarray |
Bucket 4 of 8: arg 1: rank 1, kind float, dtype float32
1 implementation from 1 paper share this bucket (rank, kind, dtype of each array argument, positional; each member's recorded signature, argument name included, is shown under it); 1 distinct output, largest cluster first. Values are the first 8 of the recorded output, flattened.
| Cluster (same digest to six decimals) | Members | Shared output on this bucket's input |
|---|---|---|
| 1 implementation 1 paper 0c77544de6fb |
src/models/_A2C_discrete.py 9559ea8ea49d3010 recorded z:1/float/float32; beta=2.0 |
[0.0875577, 0.14883, 0.215519, 0.0939858, 0.125979, 0.141282, 0.060948, 0.125898] shape [8] · float32 · Tensor |
Bucket 5 of 8: arg 1: rank 2, kind float, dtype float32 · arg 2: rank 1, kind float, dtype float32
1 implementation from 1 paper share this bucket (rank, kind, dtype of each array argument, positional; each member's recorded signature, argument name included, is shown under it); 1 distinct output, largest cluster first. Values are the first 8 of the recorded output, flattened.
| Cluster (same digest to six decimals) | Members | Shared output on this bucket's input |
|---|---|---|
| 1 implementation 1 paper 499238646790 |
abstention/calibration.py 1a0b58e25479df38 recorded preact:2/float/float32, biases:1/float/float32; temp=1.0 |
[0.267162, 0.0274529, 0.429569, 0.0254326, 0.0369665, 0.0409224, 0.0110002, 0.161495, …] shape [4, 8] · float32 · ndarray |
Bucket 6 of 8: arg 1: rank 2, kind float, dtype float32 · arg 2: rank 2, kind float, dtype float32
1 implementation from 1 paper share this bucket (rank, kind, dtype of each array argument, positional; each member's recorded signature, argument name included, is shown under it); 1 distinct output, largest cluster first. Values are the first 8 of the recorded output, flattened.
| Cluster (same digest to six decimals) | Members | Shared output on this bucket's input |
|---|---|---|
| 1 implementation 1 paper feda954aa726 |
graph_transformer/graph_transformer_model.py 2f6500993366c0f7 recorded x:2/float/float32, adjacency:2/float/float32; dim=-1 |
[0.99728, -0.030178, 0.0698943, -0.0315789, -0.033385, -0.0335388, -0.0311575, 0.0926643, …] shape [4, 8] · float32 · Tensor |
Bucket 7 of 8: arg 1: rank 3, kind float, dtype float32
1 implementation from 1 paper share this bucket (rank, kind, dtype of each array argument, positional; each member's recorded signature, argument name included, is shown under it); 1 distinct output, largest cluster first. Values are the first 8 of the recorded output, flattened.
| Cluster (same digest to six decimals) | Members | Shared output on this bucket's input |
|---|---|---|
| 1 implementation 1 paper e53c681093b7 |
capsule_network.py d55eaf6b69be6376 recorded input:3/float/float32; dim=1 |
[0.0423078, 0.57328, 0.135348, 0.186576, 0.264771, 0.0312582, 0.241622, 0.0617384, …] shape [2, 4, 8] · float32 · Tensor |
Bucket 8 of 8: arg 1: rank 4, kind float, dtype float32
1 implementation from 1 paper share this bucket (rank, kind, dtype of each array argument, positional; each member's recorded signature, argument name included, is shown under it); 1 distinct output, largest cluster first. Values are the first 8 of the recorded output, flattened.
| Cluster (same digest to six decimals) | Members | Shared output on this bucket's input |
|---|---|---|
| 1 implementation 1 paper e318343bfebd |
scripts-predict/predict.py 442a2012d75c4821 recorded logits:4/float/float32 |
[0.183535, 0.171122, 0.378186, 0.267158, 0.445652, 0.117016, 0.108935, 0.328396, …] shape [2, 3, 4, 4] · float64 · ndarray |
Identical values to six decimals (the recorded digest) on the shared input are agreement on those inputs, not equivalence; where a cluster's members carry recorded values, the largest difference among them is shown under the cluster. Paper titles are the archive's archive 2025-07-28 where the paper is in the archive and the graph's where it was added by Syntology; papers with no page here are shown by their recorded paper id only. A paper count above the implementation count means one implementation (one code sha) is held from several papers' repositories and counts once. Per-sample status, licence and fingerprint records for each paper are on its paper page. JSON twin: /census/softmax.json.
Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections