Every commit scored for change-risk against this repo's own history, so 'elevated' means elevated here rather than on some global curve.
Needs review
156 commits sit in this repo's top risk tercile, which is 33% of the 470scored. The cut is drawn against this codebase's own history rather than a global curve, so a quiet repo still fills its top band, and here it starts at 7.6 out of 10. What pushes a commit up is size and spread together: a large change confined to one area scores below a smaller one scattered across a dozen files.
Commit categories over time, read off the subject line. Fixes carry the accent because that is the series this chart exists to show.
Began feature-led, now leaning other.
Ranked by change-risk, highest first. Priority is a tercile of this repo's own distribution, so a quiet repo still fills its top band.
| # | Commit | Author | When | Lines | Risk | Top driver |
|---|---|---|---|---|---|---|
| 1 | 5da89254[Docs] Update README and leaderboards layout (#1008) | ioangatop | 4mo ago | +7.0K -5.6K | 100%Elevated | more lines added than baseline |
| 2 | b86bd547Unify prompting and answer extraction logic and introduce factories (#932) | Max Emanuel Feucht | 9mo ago | +3.1K -284 | 99%Elevated | more lines added than baseline |
| 3 | 8fb71f7cAdd `eva.multimodal` (#874) | Nicolas Känzig | 11mo ago | +5.1K -445 | 99%Elevated | more lines added than baseline |
| 4 | aac0bda5add midnight-12k to leaderboard (#799) | Nicolas Känzig | 1y ago | +3.4K -2.9K | 99%Elevated | more lines added than baseline |
| 5 | f5d80152[Radiology] Update the radiology leaderboard (#1021) | ioangatop | 2mo ago | +2.2K -799 | 99%Elevated | more lines added than baseline |
| 6 | 3460435eAdd `eva.language` module (#831) | Rita Kurban | 1y ago | +2.7K -203 | 99%Elevated | more lines added than baseline |
| 7 | 5510f6b5Update leaderboard style to heatmap (#631) | roman807 | 1y ago | +2.6K -14 | 99%Elevated | more lines added than baseline |
| 8 | 20da6446Add support for WSI-level classification (#542) | roman807 | 2y ago | +3.6K -235 | 99%Elevated | more lines added than baseline |
| 9 | 0e46864fadded boxed formatting (#973) | jonaserb-k | 8mo ago | +1.1K -15 | 98%Elevated | more lines added than baseline |
| 10 | c36237c7Increase patience to 20% and increase `n_runs` to 20 for slide level classification (#772) | Nicolas Känzig | 1y ago | +3.3K -2.1K | 98%Elevated | more lines added than baseline |
| 11 | 4a479eb2update leaderboard (#711) | Nicolas Känzig | 1y ago | +1.8K -1.5K | 98%Elevated | more lines added than baseline |
| 12 | 36c1b4e7Support semantic segmentation downstream evaluation tasks (#517) | ioangatop | 2y ago | +2.3K -131 | 98%Elevated | more lines added than baseline |
| 13 | fc83fc51Add gitleaks to CI workflow (#357) | a-thiery | 2y ago | +1.6K -0 | 98%Elevated | more lines added than baseline |
| 14 | 5710f2dfAdd `radiology` leaderboard (#856) | Nicolas Känzig | 10mo ago | +1.4K -97 | 97%Elevated | more lines added than baseline |
| 15 | bdc072ffUpdate Leaderboard (#670) | Nicolas Känzig | 1y ago | +1.7K -982 | 97%Elevated | more lines added than baseline |
| 16 | b3414772 [radiology] Add `KiTS23`, `FLARE22` and `WORD` segmentation CT radiology datasets (#1006) | ioangatop | 4mo ago | +1.2K -12 | 96%Elevated | more lines added than baseline |
| 17 | e6e46425Add `VllmModel` wrapper to `eva.multimodal` (#958) | Nicolas Känzig | 8mo ago | +898 -256 | 96%Elevated | more lines added than baseline |
| 18 | 0aea790eAdd `2D` radiology configs (#845) | Nicolas Känzig | 1y ago | +641 -0 | 96%Elevated | more lines added than baseline |
| 19 | 0d63dc7bAdd `3D` data transforms (#787) | Nicolas Känzig | 1y ago | +843 -313 | 96%Elevated | more lines added than baseline |
| 20 | e192eb9fAdd starplot leaderboard script to tools (#682) | roman807 | 1y ago | +904 -797 | 96%Elevated | more lines added than baseline |
| 21 | 55f52585Add model registry for accessing backbones by name (#591) | Nicolas Känzig | 2y ago | +1.1K -1.6K | 96%Elevated | more lines added than baseline |
| 22 | 14f95cebAdd core data modules (#17) | ioangatop | 2y ago | +1.2K -11 | 96%Elevated | more lines added than baseline |
| 23 | 94f10347Add package and dependency manager (#2) | ioangatop | 2y ago | +954 -2 | 96%Elevated | more lines added than baseline |
| 24 | 987e0919[Radiology] Add ct scan level classification datasets (#1010) | ioangatop | 3mo ago | +991 -16 | 95%Elevated | more lines added than baseline |
| 25 | 3856aed6Update .lock file to solve `exceptiongroup` dependency issue in CI (#605) | Nicolas Känzig | 2y ago | +729 -670 | 95%Elevated | more lines added than baseline |
| 26 | e7920e5dAdd `BCSS` dataset (#559) | Nicolas Känzig | 2y ago | +907 -41 | 95%Elevated | more lines added than baseline |
| 27 | 9b09117f[Radiology] Add SuPreM and Universal radiology backbones. (#993) | ioangatop | 6mo ago | +435 -2 | 93%Elevated | more lines added than baseline |
| 28 | c098904dAdding Path MMU Atlas to Eva and adding download utilites (#980) | jonaserb-k | 6mo ago | +719 -7 | 93%Elevated | more lines added than baseline |
| 29 | 3f68827bAdd `StratifiedRandomSampler` to class samplers (#916) | Max Emanuel Feucht | 10mo ago | +408 -85 | 93%Elevated | more lines added than baseline |
| 30 | 7a750aa3Add `G-Eval` LLM Judge (#914) | Nicolas Känzig | 10mo ago | +611 -46 | 93%Elevated | more lines added than baseline |
| 31 | 1f44cfc7Add JSON answer prompt template for multiple choice tasks (#905) | Nicolas Känzig | 10mo ago | +772 -127 | 93%Elevated | more lines added than baseline |
| 32 | 25ac48f1Add integration tests for segmentation tasks (#833) | Nicolas Känzig | 1y ago | +564 -6 | 93%Elevated | more lines added than baseline |
| 33 | b783adbcAdd `MSDTask7Pancreas` dataset (#832) | Nicolas Känzig | 1y ago | +864 -8 | 93%Elevated | more lines added than baseline |
| 34 | e347653fUpdate `LiTS17` dataset to make it compatible with 3D pipelines (#830) | Nicolas Känzig | 1y ago | +744 -992 | 93%Elevated | more lines added than baseline |
| 35 | 0c37a1aeAdd `VoCo` to model registry (#798) | Nicolas Känzig | 1y ago | +680 -23 | 93%Elevated | more lines added than baseline |
| 36 | baf69f47Add `BTCV` dataset / support for 3D radiology datasets (#785) | Nicolas Känzig | 1y ago | +679 -309 | 93%Elevated | more lines added than baseline |
| 37 | c984e5f3Add support for offline total segmentator training (#504) | ioangatop | 2y ago | +653 -111 | 93%Elevated | more lines added than baseline |
| 38 | 9d998879Added configs for `owkin/phikon` model (#203) | Nicolas Känzig | 2y ago | +334 -0 | 93%Elevated | more lines added than baseline |
| 39 | 981ae103Add `QuiltVQA` dataset & `FreeFormQuestionPromptTemplate` template (#912) | Nicolas Känzig | 10mo ago | +563 -88 | 91%Elevated | more lines added than baseline |
| 40 | cdf1bc43Add `CoNSeP` dataset with patching (#547) | Nicolas Känzig | 2y ago | +738 -108 | 91%Elevated | more lines added than baseline |
| 41 | 3d55c57bAdd `MultiEmbeddingsClassificationDataset` (#351) | Nicolas Känzig | 2y ago | +595 -69 | 91%Elevated | more lines added than baseline |
| 42 | e217cbe6Add `CRC_HE_NONORM` dataset (#183) | ioangatop | 2y ago | +598 -72 | 91%Elevated | more lines added than baseline |
| 43 | f2c89a90Add offline workflows for `BACH` and `PatchCamelyon` (#135) | Nicolas Känzig | 2y ago | +520 -50 | 91%Elevated | more lines added than baseline |
| 44 | d3646694Add core metrics module (#30) | ioangatop | 2y ago | +408 -2 | 91%Elevated | more lines added than baseline |
| 45 | 07078728Add record_datasets_as_runs setting to trainer (#996) | Nicolas Känzig | 3mo ago | +415 -9 | 90%Elevated | more lines added than baseline |
| 46 | 02827261Align `HuggingFaceModel` in `eva.language` with `multimodal` (#961) | Nicolas Känzig | 8mo ago | +375 -159 | 90%Elevated | more lines added than baseline |
| 47 | 435b3775Add raw answer prompt template for multiple choice tasks (#911) | Elias Baumann | 10mo ago | +359 -0 | 90%Elevated | more lines added than baseline |
| 48 | 936e96d0Add support for dataloader samplers (#713) | Nicolas Känzig | 1y ago | +389 -29 | 90%Elevated | more lines added than baseline |
| 49 | ed73bb06Add `LiTSBalanced` dataset (#648) | Nicolas Känzig | 1y ago | +418 -4 | 90%Elevated | more lines added than baseline |
| 50 | 58e54b6eAdd `LiTS` radiology ct segmentation dataset and task (#567) | ioangatop | 2y ago | +656 -13 | 90%Elevated | more lines added than baseline |
Two views of the same model: where the cuts fall, and what commit shape lands you above them.
Every scored commit, binned on the raw 0 to 10 score rather than the percentile. Percentile ranks are uniform by construction, so that axis has no shape to draw. The dashed lines are the tercile cuts behind each row's priority pill.
The 200 most recent commits, on their own recency sample rather than the feed above: that defaults to risk-sorted, so reusing it would plot only the top tercile and call it the spread. Big and scattered is what the model penalises. Click a dot to open it.
kaiko-ai/eva has 470 commits in its history from 18 contributors, the first of them Jan 16, 2024. In the last 90 days 1 file was touched, 1 times in total, most often generate_leaderboard_plots.py. Every commit is scored for change risk from its size, spread and the history of the files it touches.
