[Feat] Added pruning strategy from https://arxiv.org/abs/2511.16980

docs/3dgut.md

@@ -12,27 +12,25 @@ Here are the instructions on how to use this feature.
 
 #### Training
 
-Simplly passing in `--with_ut --with_eval3d` to the `simple_trainer.py` arg list will enable training with 3DGUT! And note in gsplat we only support MCMC densification strategy for 3DGUT:
+Simply passing in `--with_ut --with_eval3d` to the `simple_trainer.py` arg list will enable training with 3DGUT:
 
 ```
-python examples/simple_trainer.py mcmc --with_ut --with_eval3d ... <OTHER ARGS>
+python examples/simple_trainer.py default --with_ut --with_eval3d ... <OTHER ARGS>
 ```
 
-For benchmarking on MipNeRF360 Dataset, please checkout `examples/benchmarks/3dgut/mcmc.sh`
-
 Note if you are not familiar with how to get started with `simple_trainer.py`, please checkout [README.md](README.md) first!
 
 #### Rendering
 
 Once trained, you could view the 3DGS and play with the distortion effect supported through 3DGUT via our viewer:
 
 ```bash
-CUDA_VISIBLE_DEVICES=0 python simple_viewer_3dgut.py --ckpt results/benchmark_mcmc_1M_3dgut/garden/ckpt_29999_rank0.pt 
+CUDA_VISIBLE_DEVICES=0 python simple_viewer_3dgut.py --ckpt <RESULT_DIR>/garden/ckpt_29999_rank0.pt 
 ```
 
 Or a more comprehensive nerfstudio-style viewer to export videos. (note changing distortion is not yet supported in this comprehensive viewer!)
 ```bash
-CUDA_VISIBLE_DEVICES=0 python simple_viewer.py --with_ut --with_eval3d --ckpt results/benchmark_mcmc_1M_3dgut/garden/ckpt_29999_rank0.pt 
+CUDA_VISIBLE_DEVICES=0 python simple_viewer.py --with_ut --with_eval3d --ckpt <RESULT_DIR>/garden/ckpt_29999_rank0.pt 
 ```
 
 ### For users using gsplat' API:

docs/source/apis/strategy.rst

@@ -72,5 +72,5 @@ Below are the strategies that are currently implemented in `gsplat`:
 .. autoclass:: DefaultStrategy
     :members:
 
-.. autoclass:: MCMCStrategy
+.. autoclass:: ImprovedStrategy
     :members:

docs/source/index.rst

@@ -27,7 +27,7 @@ faster, more memory efficient, and with a growing list of new features!
 * *gsplat* is equipped with the **latest and greatest** 3D Gaussian Splatting techniques, 
   including `absgrad <https://ty424.github.io/AbsGS.github.io/>`_, 
   `anti-aliasing <https://niujinshuchong.github.io/mip-splatting/>`_,
-  `3DGS-MCMC <https://ubc-vision.github.io/3dgs-mcmc/>`_ etc. And more to come!
+  and improved densification heuristics. And more to come!
 
 
 .. raw:: html

docs/source/tests/eval.rst

@@ -28,7 +28,7 @@ Powered by `gsplat`'s efficient CUDA implementation, the training takes up to
 
 Feature Ablation
 ----------------------------------------------
-Evaluation of features provided in `gsplat` on Mip-NeRF (averaged over 7 scenes). We ablate `gsplat` with default settings, with absgrad and mcmc densification strategies, and antialiased mode.
+Evaluation of features provided in `gsplat` on Mip-NeRF (averaged over 7 scenes). We ablate `gsplat` with default settings, the absgrad densification heuristic, and antialiased mode.
 Absgrad method uses `--grow_grad2d 0.0006` config. These results are obtained with an A100.
 
 +-----------------------------+-------+-------+-------+----------+---------+------------+
@@ -40,16 +40,8 @@ Absgrad method uses `--grow_grad2d 0.0006` config. These results are obtained wi
 +-----------------------------+-------+-------+-------+----------+---------+------------+
 | antialiased                 | 29.03 | 0.87  | 0.14  | 3377807  | 5.87    | 19.52      |
 +-----------------------------+-------+-------+-------+----------+---------+------------+
-| mcmc  (1 mill)              | 29.18 | 0.87  | 0.14  | 1000000  | 1.98    | 15.42      |
-+-----------------------------+-------+-------+-------+----------+---------+------------+
-| mcmc  (2 mill)              | 29.53 | 0.88  | 0.13  | 2000000  | 3.43    | 21.79      |
-+-----------------------------+-------+-------+-------+----------+---------+------------+
-| mcmc  (3 mill)              | 29.65 | 0.89  | 0.12  | 3000000  | 4.99    | 27.63      |
-+-----------------------------+-------+-------+-------+----------+---------+------------+
 | absgrad & antialiased       | 29.14 | 0.88  | 0.13  | 2563156  | 4.57    | 18.43      |
 +-----------------------------+-------+-------+-------+----------+---------+------------+
-| mcmc & antialiased          | 29.23 | 0.87  | 0.14  | 1000000  | 2.00    | 15.75      |
-+-----------------------------+-------+-------+-------+----------+---------+------------+
 
 
 Trains Faster with Less GPU Memory
@@ -239,4 +231,4 @@ The evaluation of `inria-X` can be
 reproduced with our forked wersion of the official implementation at 
 `here <https://github.com/hbb1/diff-surfel-rasterization>`_;
 you need to change the :code:`--model_type 2dgs` to :code:`--model_type 2dgs-inria` in
-:code:`benchmars/basic_2dgs` and run command :code:`cd examples; bash benchmarks/basic_2dgs.sh` (commit 28c928a).
+:code:`benchmars/basic_2dgs` and run command :code:`cd examples; bash benchmarks/basic_2dgs.sh` (commit 28c928a).