Papers › Exposing flaws of generative model evaluation metrics and their unfair treatment of...
Exposing flaws of generative model evaluation metrics and their unfair treatment of diffusion models
George Stein, Jesse C. Cresswell, Rasa Hosseinzadeh, Yi Sui, Brendan Leigh Ross, Valentin Villecroze, Zhaoyan Liu, Anthony L. Caterini, J. Eric T. Taylor, Gabriel Loaiza-Ganem
We systematically study a wide variety of generative models spanning semantically-diverse image datasets to understand and improve the feature extractors and metrics used to evaluate them. Using best practices in psychophysics, we measure human perception of image realism for generated samples by conducting the largest experiment evaluating generative models to date, and find that no existing metric strongly correlates with human evaluations. Comparing to 17 modern metrics for evaluating the overall performance, fidelity, diversity, rarity, and memorization of generative models, we find that the state-of-the-art perceptual realism of diffusion models as judged by humans is not reflected in commonly reported metrics such as FID. This discrepancy is not explained by diversity in generated samples, though one cause is over-reliance on Inception-V3. We address these flaws through a study of alternative self-supervised feature extractors, find that the semantic information encoded by individual networks strongly depends on their training procedure, and show that DINOv2-ViT-L/14 allows for much richer evaluation of generative models. Next, we investigate data memorization, and find that generative models do memorize training examples on simple, smaller datasets like CIFAR10, but not necessarily on more complex datasets like ImageNet. However, our experiments show that current metrics do not properly detect memorization: none in the literature is able to separate memorization from other phenomena such as underfitting or mode shrinkage. To facilitate further development of generative models and their evaluation we release all generated image datasets, human evaluation data, and a modular library to compute 17 common metrics for 9 different encoders at https://github.com/layer6ai-labs/dgm-eval.
In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.
Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.
For agents, on Syntology's MCP service (how to connect):
get_harvested_code_for_paper(arxiv_id="2306.04675")get_code(code_sha256="…")with a sample'scode_sha256from the list belowget_citation_path(paper_1="2306.04675", paper_2="…")with another paper's arXiv id or titleget_concepts_for_paper(arxiv_id="2306.04675")
Code
Syntology 63 code samples: 62 from 8 repositories linked to this paper, 1 identical to code harvested from another paper's repository. Across all 63: 53 ran (of which 0 constructed an object rather than computing a result; 30 with no instrument failure: 6 honoured, 1 violated, 23 with no contract checked; 23 where Syntology's instrument failed) · 10 unverified. Where Syntology's instrument failed: 18 where our contract draft was wrong; 5 where the contract check could not run; failures of Syntology's instrument, not of the code.
By repository: official repository: 3 samples from 1 repository, 2 ran (0 with no instrument failure; 2 where Syntology's instrument failed); found in paper text by Syntology: 59 samples from 7 repositories, 50 ran (30 with no instrument failure; 20 where Syntology's instrument failed); 1 identical to code first harvested elsewhere. The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.
No sample is linked to this paper from 13 of its repositories that Syntology harvested for other papers:
- layer6ai-labs/dgm_manifold_survey, community (archive-listed): harvested for another paper, listed under the repository
- louaaron/scaling-riemannian-diffusion, community (archive-listed): harvested for another paper, listed under the repository
- openai/improved-diffusion, found in paper text by Syntology: Syntology's graph links these samples to Analytic-DPM: an Analytic Estimate of the Optimal Reverse Variance in Diffusion... (arXiv:2201.06503; that paper's own run record: 2 ran (of which 1 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified); nothing checks that they implement this paper's method.
- NVlabs/stylegan2-ada-pytorch, found in paper text by Syntology: Syntology's graph links these samples to Training Generative Adversarial Networks with Limited Data (arXiv:2006.06676; that paper's own run record: 21 ran (of which 0 constructed an object rather than computing a result; 19 with no instrument failure: 2 honoured, 1 violated, 16 with no contract checked; 2 where Syntology's instrument failed) · 8 unverified); nothing checks that they implement this paper's method.
- NVlabs/ffhq-dataset, found in paper text by Syntology: Syntology's graph links these samples to Local MAP Sampling for Diffusion Models (arXiv:2510.07343; that paper's own run record: 5 ran (of which 2 constructed an object rather than computing a result; 2 with no instrument failure: 0 honoured, 0 violated, 2 with no contract checked; 3 where Syntology's instrument failed) · 2 unverified); nothing checks that they implement this paper's method.
- NVlabs/LSGM, found in paper text by Syntology: Syntology's graph links these samples to Score-based Generative Modeling in Latent Space (arXiv:2106.05931; that paper's own run record: 2 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 2 where Syntology's instrument failed) · 3 unverified); nothing checks that they implement this paper's method.
- newbeeer/pfgmpp, found in paper text by Syntology: Syntology's graph links these samples to PFGM++: Unlocking the Potential of Physics-Inspired Generative Models (arXiv:2302.04265; that paper's own run record: 2 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 1 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 1 unverified); nothing checks that they implement this paper's method.
- autonomousvision/stylegan-xl, found in paper text by Syntology: Syntology's graph links these samples to Provably Near-Optimal Federated Ensemble Distillation with Negligible Overhead (arXiv:2502.06349; that paper's own run record: 19 ran (of which 0 constructed an object rather than computing a result; 6 with no instrument failure: 5 honoured, 0 violated, 1 with no contract checked; 13 where Syntology's instrument failed) · 3 unverified); nothing checks that they implement this paper's method.
- CompVis/latent-diffusion, found in paper text by Syntology: Syntology's graph links these samples to Pseudo Numerical Methods for Diffusion Models on Manifolds (arXiv:2202.09778; that paper's own run record: 10 ran (of which 4 constructed an object rather than computing a result; 4 with no instrument failure: 0 honoured, 0 violated, 4 with no contract checked; 6 where Syntology's instrument failed) · 4 unverified) and AI for Cultural Heritage Textiles: Fine-Tuned Latent Diffusion for Novel Ulos Motif Synthesis (arXiv:2607.06590; that paper's own run record: 11 ran (of which 0 constructed an object rather than computing a result; 5 with no instrument failure: 0 honoured, 5 violated, 0 with no contract checked; 6 where Syntology's instrument failed) · 2 unverified); nothing checks that they implement this paper's method.
- microsoft/StyleSwin, found in paper text by Syntology: Syntology's graph links these samples to StyleSwin: Transformer-based GAN for High-resolution Image Generation (arXiv:2112.10762; that paper's own run record: 5 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 5 where Syntology's instrument failed) · 0 unverified); nothing checks that they implement this paper's method.
- samb-t/unleashing-transformers, found in paper text by Syntology: Syntology's graph links these samples to Structured Denoising Diffusion Models in Discrete State-Spaces (arXiv:2107.03006; that paper's own run record: 5 ran (of which 4 constructed an object rather than computing a result; 4 with no instrument failure: 0 honoured, 0 violated, 4 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified); nothing checks that they implement this paper's method.
- casey-meehan/data-copying, found in paper text by Syntology: Syntology's graph links these samples to A Non-Parametric Test to Detect Data-Copying in Generative Models (arXiv:2004.05675; that paper's own run record: 4 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 1 violated, 0 with no contract checked; 3 where Syntology's instrument failed) · 2 unverified); nothing checks that they implement this paper's method.
- eyalbetzalel/fcd, found in paper text by Syntology: Syntology's graph links these samples to A Study on the Evaluation of Generative Models (arXiv:2206.10935; that paper's own run record: 2 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 1 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 1 unverified); nothing checks that they implement this paper's method.
Levels written between 2026-09-22 and 2026-09-26 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed. 14 of the 53 samples that ran have no recorded date.
Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.
A paper named beside a repository with no sample linked to this paper is shown with that paper's own run record, not this paper's: “ran” means executed on a synthesized input, not that the code is correct, and “where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Papers are listed in arXiv-id order, at most three per repository.
Code Syntology ran Syntology
63 samples harvested: 53 ran (of which 0 constructed an object rather than computing a result; 30 with no instrument failure: 6 honoured, 1 violated, 23 with no contract checked; 23 where Syntology's instrument failed) · 10 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page last changed ; when this build read Syntology's graph is in the build record. That date is not when the samples ran.
Licence: 12 of the 63 samples are pointer only, meaning Syntology does not serve that copy's text. For 11 of them, this paper's copy is under a permissive licence and cleared, but the stored copy of the same code that Syntology serves is not; the licence label's tooltip gives Syntology's recorded reason. This page shows no code text for any sample; each one links to its file in the repository.
Harvested from 8 repositories linked to this paper, official or community; the list below is grouped by repository, each group headed by its name and what it is to the paper. 1 sample is listed last, in a group without a repository. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.
Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.
Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: 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. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.
A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity. A census tag appears when Syntology's census ran implementations with that name on one shared input per bucket: it gives how many were compared, from how many papers, and how many distinct outputs they produced, and names this sample's bucket when this sample was one of them.
de1462c410a454f6 · report
fixture: returned list · level recorded 2026-09-22 0e74fdd296ca6e73 · report
fixture: signature device:NoneType/0/?, num_workers:NoneType/0/? · returned tuple · level recorded 2026-09-22 2bbe398e90e76c2d · report
3e0a8dc623475d8e · report
fixture: signature ns:Tensor/2/float, p:Tensor/2/float · returned Tensor · level recorded, date not recorded 9b16691c6712e596 · report
fixture: signature g:Tensor/2/float, x:Tensor/2/float, n_power_series:int/0/int, vareps:Tensor/2/float, coeff_fn:callable/0/?, training:bool/0/bool · returned Tensor · level recorded 2026-09-22 85a4b5aac95f359d · report
fixture: signature g:Tensor/2/float, x:Tensor/2/float · returned Tensor · level recorded 2026-09-22 70b3429a213d3584 · report
fixture: signature M:Tensor/3/float · returned Tensor · level recorded 2026-09-22 f6b944f50d3f15ae · report
fixture: signature model:callable/0/? · returned int · level recorded, date not recorded 2dc23e3f121c6a38 · report
fixture: signature z:Tensor/2/float · returned Tensor · level recorded, date not recorded 2168d015a2e457cb · report
fixture: signature size:tuple/1/int · returned Tensor · level recorded, date not recorded 1b2e0a7656cc76a7 · report
fixture: signature one_hot_targets:ArrayImpl/2/float, label_smoothing:float/0/float · returned ArrayImpl · level recorded 2026-09-26 ffc38a27fd9393ba · report
fixture: signature h:MockHyperparameter/0/? · returned list · level recorded 2026-09-24 7dfd4d71b4b83843 · report
fixture: signature h:HyperparameterConfig/0/? · returned dict · level recorded 2026-09-24 cd108d44d16c6e3c · report
fixture: signature h:MockHyperObject/0/? · returned dict · level recorded 2026-09-26 1983d19e3e1ab60d · report
fixture: signature train:bool/0/bool, dtype:callable/0/?, norm_type:str/0/str · returned partial · level recorded 2026-09-26 census: 3 implementations from 7 papers, 3 distinct outputs across 3 buckets, one shared input per bucket maskgit/libml/losses.py ran fingerprinted Apache-2.0 (permissive) ·
06f87811f5c2b96a · report
fixture: signature y_true:ArrayImpl/2/float, y_pred:ArrayImpl/2/float · returned ArrayImpl · level recorded 2026-09-26 95c51644bf3cb9ff · report
fixture: signature path:str/0/str · returned dict · level recorded 2026-09-26 877afab065f859ba · report
fixture: signature ratio:float/0/float, total_unknown:int/0/int, method:str/0/str · returned ArrayImpl · level recorded 2026-09-26 d80134d1097de9c4 · report
fixture: signature a:ArrayImpl/2/float, b:ArrayImpl/2/float, b2:ArrayImpl/2/float, precision:NoneType/0/? · returned ArrayImpl · level recorded 2026-09-26 38e696a22dc81a6f · report
fixture: signature x:ArrayImpl/4/float, window_shape:tuple/1/int, strides:tuple/1/int, padding:str/0/str · returned ArrayImpl · level recorded 2026-09-26 0408240381821d2c · report
fixture: signature stddev:float/0/float, dtype:callable/0/? · returned function · level recorded 2026-09-26 census: 11 implementations from 16 papers, 9 distinct outputs across 3 buckets, one shared input per bucket maskgit/nets/layers.py ran Apache-2.0 (permissive) ·
5829218c5489c6b6 · report
fixture: signature x:ArrayImpl/4/float, factor:int/0/int · returned ArrayImpl · level recorded 2026-09-26 389c62669d4417ea · report
9bd8ee32e810e900 · report
fixture: signature x:ArrayImpl/2/float, beta:float/0/float · returned ArrayImpl · level recorded 2026-09-26 8c4274c2f8994da8 · report
fixture: signature decay_scheme:str/0/str, init_lr:float/0/float, warmup_steps:int/0/int, decay_steps:int/0/int, decay_rate:float/0/float, decay_start:int/0/int, min_lr:float/0/float · returned NarrowCosineDecay · level recorded 2026-09-26 d177a0713d0bea82 · report
fixture: signature scale:float/0/float, dtype:callable/0/? · returned function · level recorded 2026-09-26 0f49a2b0642d1098 · report
fixture: signature minval:float/0/float, maxval:float/0/float, dtype:callable/0/? · returned function · level recorded 2026-09-26 025e4c7cda618f25 · report
fixture: signature s:str/0/str · returned list · level recorded 2026-09-22 6a0f2095bdd516ab · report
fixture: signature s:str/0/str · returned tuple · level recorded 2026-09-26 census: 13 implementations from 17 papers, 5 distinct outputs across 3 buckets, one shared input per bucket models/generator.py ran; the contract check could not run (RAISES) · metamorphic tier: invariant fingerprinted MIT (permissive) ·
144d10b49baeb8a6 · report
fixture: signature x:Tensor/4/float, window_size:int/0/int · returned Tensor · level recorded, date not recorded 258c1af08f41aa17 · report
fixture: signature dataset:SimpleDataset/0/?, shuffle:bool/0/bool, distributed:bool/0/bool · returned RandomSampler · level recorded 2026-09-26 9865a02804b0266f · report
fixture: signature in_channels:int/0/int · returned tuple · level recorded 2026-09-26 6f65e378a4313f87 · report
fixture: signature k:list/1/float · returned Tensor · level recorded, date not recorded 43108d105300b7f5 · report
fixture: signature image:Tensor/4/float · returned Tensor · level recorded 2026-09-26 001e240d5a8fc4dc · report
fixture: signature q0:Tensor/4/float, q1:Tensor/4/float, k0:Tensor/4/float, k1:Tensor/4/float, hw:int/0/int, ws:int/0/int, nheads:int/0/int · returned Tensor · level recorded 2026-09-26 census: 5 implementations from 7 papers, 1 distinct output on one shared input this sample: bucket 1 of 1, arg 1: rank 4, kind float, dtype float32 · arg 2: rank 2, kind float, dtype float32 op/upfirdn2d.py ran; the contract check could not run (RAISES) · metamorphic tier: invariant MIT (permissive) here; the stored copy is pointer only ·
fe99cfd294676edb · report
fixture: signature input:Tensor/4/float, kernel:Tensor/2/float, up_x:int/0/int, up_y:int/0/int, down_x:int/0/int, down_y:int/0/int, pad_x0:int/0/int, pad_x1:int/0/int, pad_y0:int/0/int, pad_y1:int/0/int · returned Tensor · level recorded 2026-09-22 61bf152e6a42a184 · report
fixture: signature windows:Tensor/4/float, window_size:int/0/int, H:int/0/int, W:int/0/int · returned Tensor · level recorded, date not recorded 12b482fc759f87dd · report
0def50a351111624 · report
c1f9693596855557 · report
9cd07419f7c926f8 · report
fixture: signature s:str/0/str · returned tuple · level recorded, date not recorded census: 7 implementations from 5 papers, 5 distinct outputs across 2 buckets, one shared input per bucket diffusion-projected-gan/pg_modules/diffaug.py ran fingerprinted MIT (permissive) ·
da4e9834e767ce2d · report
fixture: signature x:Tensor/4/float, policy:str/0/str, channels_first:bool/0/bool · returned Tensor · level recorded 2026-09-26 ef1e7dffb5d433d4 · report
fixture: signature c:int/0/int, mode:str/0/str · returned BatchNorm2d · level recorded 2026-09-22 f87cd99314206f72 · report
fixture: signature in_planes:int/0/int, out_planes:int/0/int · returned Sequential · level recorded 2026-09-24 0186912dcb8b92b9 · report
fixture: signature in_planes:int/0/int, out_planes:int/0/int · returned Sequential · level recorded 2026-09-24 a2b45afa097b55b6 · report
fixture: signature name:str/0/str · returned str · level recorded, date not recorded 053fc534bc6bb989 · report
fixture: signature seconds:float/0/float · returned str · level recorded 2026-09-25 77f4aa0649e7f404 · report
fixture: signature seconds:float/0/float · returned str · level recorded 2026-09-25 456c172851946697 · report
fixture: signature img:Tensor/4/float, grid_w:NoneType/0/?, grid_h:int/0/int, float_to_uint8:bool/0/bool, chw_to_hwc:bool/0/bool, to_numpy:bool/0/bool · returned ndarray · level recorded 2026-09-22 f47dff4b7b27bba7 · report
fixture: signature translate:tuple/1/float, angle:float/0/float · returned ndarray · level recorded 2026-09-22 dd1dc700e87f36cc · report
fixture: signature a:int/0/int, b:NoneType/0/? · returned int · level recorded, date not recorded 02870cd90e053ab0 · report
fixture: signature x:Tensor/2/float, dim:int/0/int, eps:float/0/float · returned Tensor · level recorded 2026-09-26 d01b68a634f71551 · report
fixture: signature s:str/0/str · returned list · level recorded, date not recorded 236de8614a178382 · report
fixture: signature s:str/0/str · returned list · level recorded 2026-09-22 180016b9227381b1 · report
fixture: signature s:str/0/str · returned tuple · level recorded 2026-09-22 census: 7 implementations from 9 papers, 5 distinct outputs across 2 buckets, one shared input per bucket this sample: bucket 1 of 2, arg 1: rank 4, kind float, dtype float32 diffusion-projected-gan/pg_modules/diffaug.py ran · our draft was wrong fingerprinted MIT (permissive) here; the stored copy is pointer only ·
011230b2b9b8fb6f · report
fixture: signature x:Tensor/4/float · returned Tensor · level recorded, date not recorded census: 6 implementations from 9 papers, 3 distinct outputs on one shared input this sample: bucket 1 of 1, arg 1: rank 4, kind float, dtype float32 diffusion-projected-gan/pg_modules/diffaug.py ran · our draft was wrong fingerprinted MIT (permissive) here; the stored copy is pointer only ·
5b0d8787e670fc63 · report
fixture: signature x:Tensor/4/float · returned Tensor · level recorded, date not recorded 9d31d2c4cd16bb2d · report
fc49fa553c005268 · report
6d3c813b7561493d · report
1 sample is identical to code Syntology harvested from another paper's repository; this paper's own copy was not located, so no repository is named and no licence is recorded.
8241c0562bc710fd · report
fixture: signature it:list/1/int, size:int/0/int · returned callable_iterator · level recorded, date not recorded Tasks
Results from the paper archive 2025-07-28
Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.
Methods
1 archive method tag without a method page not shown.
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