Polish dataset card with detailed, paper-grounded documentation

- Replace placeholder data schema with accurate modalities, per-step
state, and on-disk episode format
- Add simulation scenarios, contact-splat encoding, and data-generation
(Project Chrono + Ornstein-Uhlenbeck) details
- Document repository layout, data splits, selective downloading,
intended uses, and limitations
- Summarize the associated ChronoDreamer model and VLM-AUC protocol
- Keep citation as a neutral placeholder (paper pending)

README.md

@@ -18,96 +18,74 @@ tags:
 - simulator
 - friction
 - contact-dynamics
+- contact-mechanics
 - physics-simulation
 - dynamics-prediction
+- robotic-manipulation
+- gaussian-splat
+- project-chrono
 pretty_name: DreamerBench
 size_categories:
 - n<1K
 ---
 
-# Dataset Card for DreamerBench
+# DreamerBench
+
+**A simulator-generated, multimodal benchmark for world models in contact-rich, frictional robotic manipulation.**
+
+DreamerBench provides time-synchronized RGB observations, camera-aligned **contact-splat** visualizations, proprioception, low-level control actions, and dense physics annotations (contact forces, friction coefficients, and contact-mode flags), together with **precomputed discrete visual latents** for token-based world models. It is the dataset used to train and evaluate **ChronoDreamer**, an action-conditioned world model for contact-rich manipulation.
+
+| | |
+|---|---|
+| **Code** | <https://github.com/uwsbel/ChronoDreamer> |
+| **Dataset** | <https://huggingface.co/datasets/zzhou292/DreamerBench> |
+| **Generator** | [Project Chrono](https://projectchrono.org/) multi-physics engine |
+| **License** | MIT |
+| **Point of contact** | Json Zhou (`zzhou292@wisc.edu`) |
+
+> [!NOTE]
+> DreamerBench is a *focused* benchmark (~12 hours of interaction data), deliberately narrowed to contact-rich tabletop dynamics rather than a broad, web-scale corpus. Its value is the explicit contact/friction supervision that general video datasets omit.
 
 ## Table of Contents
+
+- [Overview](#overview)
+  - [Highlights](#highlights)
+  - [Visual Data Samples](#visual-data-samples)
 - [Dataset Description](#dataset-description)
-  - [Dataset Summary](#dataset-summary)
-  - [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
-  - [Languages](#languages)
+  - [Simulation Scenarios](#simulation-scenarios)
+  - [Data Modalities and Per-Step State](#data-modalities-and-per-step-state)
+  - [Contact Encoding via Depth-Weighted Gaussian Splats](#contact-encoding-via-depth-weighted-gaussian-splats)
+  - [Data Generation](#data-generation)
 - [Dataset Structure](#dataset-structure)
-  - [Data Instances](#data-instances)
-  - [Data Fields](#data-fields)
+  - [Repository Layout](#repository-layout)
+  - [On-Disk Episode Format](#on-disk-episode-format)
+  - [Precomputed Latents (Cosmos DI8x8)](#precomputed-latents-cosmos-di8x8)
   - [Data Splits](#data-splits)
   - [Loading the Data](#loading-the-data)
-- [Dataset Creation](#dataset-creation)
-  - [Curation Rationale](#curation-rationale)
-  - [Source Data](#source-data)
-  - [Personal and Sensitive Information](#personal-and-sensitive-information)
-- [Considerations for Using the Data](#considerations-for-using-the-data)
-  - [Social Impact of Dataset](#social-impact-of-dataset)
-  - [Discussion of Biases](#discussion-of-biases)
-  - [Other Known Limitations](#other-known-limitations)
+- [Associated Model and Benchmark](#associated-model-and-benchmark)
+  - [ChronoDreamer World Model](#chronodreamer-world-model)
+  - [VLM-AUC Evaluation](#vlm-auc-evaluation)
+- [Intended Uses and Tasks](#intended-uses-and-tasks)
+- [Limitations](#limitations)
 - [Additional Information](#additional-information)
-  - [Dataset Curators](#dataset-curators)
-  - [Licensing Information](#licensing-information)
-  - [Citation Information](#citation-information)
-
-## Dataset Description
-
-- **Homepage:** https://huggingface.co/datasets/zzhou292/DreamerBench
-- **Repository:** https://github.com/uwsbel/ChronoDreamer
-- **Paper:** [TODO: Link to arXiv paper if available]
-- **Point of Contact:** Json Zhou, zzhou292@wisc.edu
-
-### Dataset Summary
+  - [Licensing](#licensing)
+  - [Citation](#citation)
+  - [Acknowledgments](#acknowledgments)
 
-**DreamerBench** is a large-scale dataset designed for training and evaluating **World Models** in robotics applications. Unlike standard visual-only datasets, DreamerBench explicitly focuses on physical interaction dynamics, specifically **friction** and **contact data**.
+## Overview
 
-The dataset is generated using Project Chrono (https://projectchrono.org/), simulating diverse robotic interaction scenarios where precise modeling of physical forces is critical. It includes pre-computed **encodings** to facilitate efficient training of latent dynamics models.
+World models predict how a scene evolves under an agent's actions, enabling planning by imagination. Most video-prediction datasets, however, optimize for visual plausibility and treat **contact** as an implicit, unlabeled phenomenon: friction coefficients are fixed or weakly randomized, contact forces are rarely logged, and evaluation measures task reward or perceptual quality rather than contact-mechanics accuracy. This is precisely the regime that determines whether a manipulation trajectory succeeds.
 
-Key features:
-* **Physical Fidelity:** detailed ground-truth annotations for coefficient of friction, contact forces, and slip.
-* **Multi-Modal:** Contains visual observations (RGB/Depth), proprioceptive states, and explicit physics parameters.
-* **World Model Ready:** Structured to support next-step prediction and imaginary rollout training (Dreamer-style architectures).
+DreamerBench targets this gap. It models **non-smooth hybrid dynamics** with intermittent contact, Coulomb friction, and stick–slip transitions, and exposes the underlying physical quantities as first-class, supervised signals alongside the rendered video.
 
-### Supported Tasks and Leaderboards
+### Highlights
 
-* **World Modeling / Dynamics Learning:** Training models to predict future states ($s_{t+1}$) given current state ($s_t$) and action ($a_t$).
-* **Offline Reinforcement Learning:** Learning policies from the provided simulator trajectories without active environmental interaction.
-* **Sim-to-Real Adaptation:** Using the varied friction/contact parameters to train robust policies that generalize to real-world physics.
+- **Contact-centric supervision.** Contact modes (no-contact / sticking / sliding / separating), normal and tangential forces, and friction parameters are logged with every frame and projected into camera-aligned **contact-splat** images suitable as a supervised output channel.
+- **Multimodal but tractable.** Synchronized RGB, contact splats, and proprioception are provided alongside **precomputed Cosmos DI8×8 FSQ tokens**, so models can train in pixel space or in a discrete latent space without re-running encoders.
+- **Controlled physical variation.** Friction coefficients, object inertial parameters, initial conditions, and control signals are varied across seeds within a small family of scenarios, supporting the study of generalization under physical-parameter shift.
 
-### Languages
-
-The dataset is **non-linguistic**: it consists of physics-based simulation rollouts (visual, proprioceptive, and contact/friction signals). The only natural-language content is English-language metadata, field names, and documentation (BCP-47 code: `en`).
-
-## Dataset Structure
+### Visual Data Samples
 
-### Data Instances
-
-Each instance in the dataset represents a **trajectory** or **episode** of a robot interacting with the environment.
-
-**Example structure (JSON/Parquet format):**
-
-```json
-{
-  "episode_id": "traj_001",
-  "steps": 1000,
-  "observations": {
-    "rgb": [Array of (1000, 64, 64, 3) images],
-    "depth": [Array of (1000, 64, 64, 1) images],
-    "proprioception": [Array of joint angles/velocities]
-  },
-  "actions": [Array of control inputs],
-  "rewards": [Array of float scalars],
-  "physics_data": {
-    "contact_forces": [Array of 3D force vectors],
-    "friction_coefficient": 0.8,
-    "contact_detected": [Binary array]
-  },
-  "encoding": [Pre-computed latent vectors, e.g., VAE or RSSM states]
-}
-```
-
-**Example scenarios:**
-<h3>Visual Data Samples</h3>
 <p>
   The clips below are <strong>representative, time-synchronized visualizations</strong> of four
   example scenarios. Each scenario is shown from three synchronized camera views
@@ -189,29 +167,97 @@ Each instance in the dataset represents a **trajectory** or **episode** of a rob
   </tbody>
 </table>
 
-### Data Fields
+In the contact-splat column, **red** encodes contact-force magnitude and **green–blue** encodes the projected force direction in the image plane (see [Contact Encoding](#contact-encoding-via-depth-weighted-gaussian-splats)).
+
+## Dataset Description
+
+### Simulation Scenarios
+
+DreamerBench currently covers four tabletop interaction motifs involving a hand-held tool and one or more objects. Each scenario provides synchronized ego/side RGB views, contact-splat images, proprioception, low-level actions, and physics annotations, with discrete FSQ tokens precomputed via the Cosmos DI8×8 tokenizer.
+
+| Scenario | Objects and interaction | Dominant contact behavior |
+|----------|-------------------------|---------------------------|
+| `flashlight_box` | Hand-held tool and a box on a planar surface | Sliding and pushing |
+| `flashlight_coca` | Tool and a tall cylindrical container | Sliding and rolling contact |
+| `waterbottle_coca` | Bottle interacting with a cylinder | Combined sliding and rolling along the table |
+| `fea_flashlight` | Four tetrahedron-meshed **finite-element (FEA)** deformable beams and two standing flashlights (rigid bodies) | Deformable contact and beam bending |
+
+For each scenario, multiple random seeds perturb surface friction coefficients, initial object poses and velocities, and the control trajectory, producing families of related but distinct contact sequences.
+
+### Data Modalities and Per-Step State
+
+At each simulation step `t`, DreamerBench logs a state tuple `s_t = (o_rgb, o_contact, q, q̇, f_contact, μ, c)`, together with a low-level action `a_t` and, optionally, a task reward `r_t`. An episode is the sequence `τ = (s_0, a_0, r_0, s_1, …, s_T)`. Episodes last **300–600 steps**, chosen so that objects undergo multiple contact-mode transitions (impact, stick–slip, rolling, settling).
+
+| Signal | Symbol | Description |
+|--------|:------:|-------------|
+| RGB views | `o_rgb` | Egocentric and side-camera RGB frames, rendered at **256×256**. |
+| Contact splat | `o_contact` | 256×256 camera-aligned image encoding contact-force magnitude (red) and projected direction (green/blue). |
+| Proprioception | `q`, `q̇` | Robot joint positions and velocities (`N_j = 4` joint channels). |
+| Action | `a` | 3-D end-effector velocity command `(x, y, z)`, generated by an Ornstein–Uhlenbeck process. |
+| Contact forces | `f_contact` | Per-contact 3-D normal and tangential forces (impulses discretized over Δt). |
+| Friction | `μ` | Coulomb friction coefficients (e.g., static/kinetic `(μ_s, μ_k)`), varied across seeds. |
+| Contact mode | `c` | Discrete flag: no-contact / sticking / sliding / separating. |
+| Reward *(optional)* | `r` | Scalar task reward, when a task is defined. |
+| Precomputed tokens | `z` | Cosmos DI8×8 FSQ discrete tokens (32×32 grid, vocabulary `V = 64,000`) for the RGB and contact streams. |
 
-Each episode/trajectory exposes the following fields (mirroring the schema shown above):
+### Contact Encoding via Depth-Weighted Gaussian Splats
 
-| Field | Type | Description |
-|-------|------|-------------|
-| `episode_id` | string | Unique trajectory identifier (e.g., `traj_001`). |
-| `steps` | int | Number of timesteps in the episode. |
-| `observations.rgb` | array | RGB frames, shape `(steps, 64, 64, 3)`. |
-| `observations.depth` | array | Depth frames, shape `(steps, 64, 64, 1)`. |
-| `observations.proprioception` | array | Robot joint angles / velocities per step. |
-| `actions` | array | Control inputs applied at each step. |
-| `rewards` | array | Per-step scalar rewards. |
-| `physics_data.contact_forces` | array | 3D contact-force vectors per step. |
-| `physics_data.friction_coefficient` | float | Coefficient of friction for the interaction. |
-| `physics_data.contact_detected` | array | Binary contact indicator per step. |
-| `encoding` | array | Pre-computed latent states (e.g., VAE/RSSM); see `cosmos_tokenized/` for Cosmos Tokenizer DI 8×8 latents. |
+Rather than storing contact as a sparse list of points, DreamerBench renders, at every step, a **contact-splat image** that aggregates all active contacts into a single RGB frame aligned with a reference camera — a format that vision backbones and video tokenizers can consume directly.
 
-> The inline previews above are rendered at 256×256 for display, while the stored RGB/Depth observation tensors are 64×64 as listed in the schema.
+For each contact `i` with 3-D position `p_i` and force `f_i`:
+
+1. The point is transformed into the camera frame and projected to a pixel via a pinhole model; a second point displaced along `f_i` is projected to recover the **in-plane force direction**.
+2. The **red** channel stores the clipped, normalized force magnitude; the **green/blue** channels store the normalized 2-D direction.
+3. Each contact is splatted as an isotropic Gaussian whose radius grows with force magnitude, multiplied by a **depth weight** `exp(-X_i / τ_depth)` so that nearer contacts dominate.
+4. Contributions are accumulated and per-pixel normalized, yielding a soft z-buffer over Gaussian "surfels" that compactly encodes spatial footprint, approximate pressure, and projected force direction.
+
+### Data Generation
+
+Trajectories are produced with **Project Chrono**, configured as a multi-physics engine with rigid and (optionally) flexible bodies, Coulomb friction, and penalty- or constraint-based contact handling. The continuous-time dynamics form a non-smooth hybrid ODE/DAE system; DreamerBench records a discretized version suited to sequence modeling.
+
+To excite rich, repeatable contact behavior without high-frequency jitter, the joystick command channel is driven by a three-dimensional **Ornstein–Uhlenbeck (OU)** process — temporally correlated colored noise — post-processed through a minimum-norm constraint, a deadzone, and a clipped kinematic integrator before being passed to an inverse-kinematics solver. This consistently drives sliding, rolling, and intermittent sticking while avoiding degenerate near-static trajectories, and it admits straightforward ablations (e.g., varying the correlation time or substituting i.i.d. Gaussian noise).
+
+## Dataset Structure
+
+### Repository Layout
+
+```text
+DreamerBench/
+├── flashlight_box_{00,01,02,eval}.zip     # raw episodes: RGB, contact splat, proprio, actions, physics
+├── flashlight_coca_{00,01,02,eval}.zip
+├── waterbottle_coca_{00,01,02,eval}.zip
+├── fea_flashlight_{00,01,02,eval}.zip
+├── cosmos_tokenized/                       # precomputed Cosmos DI8x8 FSQ latents
+│   ├── <scenario>_{60,120,180,eval}_tokenized.zip
+│   └── combined_{240,480,720,eval}_tokenized.zip
+├── pretrained_ckpt/                        # ChronoDreamer world-model checkpoints
+│   ├── mode_0/epoch_{0..9}.zip             # training configuration 0
+│   ├── mode_1/epoch_{0..9}.zip             # training configuration 1
+│   └── fea_flashlight/epoch_{4..7}.zip     # FEA-only ablation study
+├── inference_results.zip                   # rollouts + VLM-AUC evaluation artifacts
+├── assets/previews/                        # short MP4 preview clips (shown above)
+└── README.md
+```
+
+All archives are tracked with **Git LFS**.
+
+### On-Disk Episode Format
+
+Each scenario archive (e.g., `flashlight_box_00.zip`) contains multiple episodes and auxiliary files:
+
+- arrays for RGB and contact-splat images, proprioception, actions, and physics annotations;
+- precomputed discrete visual tokens for the RGB and contact streams (e.g., `video.bin`, `contact_splat.bin`), obtained from the Cosmos DI8×8 tokenizer via finite-scalar quantization;
+- JSON metadata describing the token grid size `S`, vocabulary size `V`, frame count, frame rate, and segmentation into episodes (segment IDs).
+
+This layout streams naturally in common training stacks (PyTorch, JAX), either as variable-length episodes or as a single concatenated sequence with associated segment IDs.
+
+### Precomputed Latents (Cosmos DI8x8)
+
+RGB and contact-splat frames are tokenized with the **NVIDIA Cosmos-0.1-Tokenizer-DI8×8**, which maps a 256×256 image to a **32×32** grid of discrete tokens (8× spatial compression per axis) via **Finite-Scalar Quantization (FSQ)**. Six latent channels are quantized with per-channel levels `(8, 8, 8, 5, 5, 5)`, giving a vocabulary of `V = 64,000` with no learnable codebook. The `cosmos_tokenized/` archives ship these latents per scenario and as combined aggregates so that token-based world models avoid repeated encoding.
 
 ### Data Splits
 
-The dataset is organized by **interaction scenario**, each provided as sharded ZIP archives plus a held-out evaluation split:
+The dataset is organized by **interaction scenario**, each provided as sharded training archives plus a held-out evaluation split:
 
 | Scenario | Training shards | Eval split |
 |----------|-----------------|------------|
@@ -220,16 +266,11 @@ The dataset is organized by **interaction scenario**, each provided as sharded Z
 | `waterbottle_coca` | `waterbottle_coca_00.zip`, `_01`, `_02` | `waterbottle_coca_eval.zip` |
 | `fea_flashlight` | `fea_flashlight_00.zip`, `_01`, `_02` | `fea_flashlight_eval.zip` |
 
-**Additional artifacts:**
-
-- **`cosmos_tokenized/`** — latent encodings produced with the **Cosmos Tokenizer DI 8×8** autoencoder. Provided per scenario (`*_60`, `*_120`, `*_180`, `*_eval`) and as combined aggregates (`combined_240`, `combined_480`, `combined_720`, `combined_eval`).
-- **`pretrained_ckpt/`** — pretrained World-Model checkpoints for use with the ChronoDreamer pipeline: `mode_0/` and `mode_1/` (`epoch_0`–`epoch_9`), plus an `fea_flashlight/` ablation (`epoch_4`–`epoch_7`).
-- **`inference_results.zip`** — sample rollout/inference outputs.
-- **`assets/previews/`** — short MP4 preview clips used in the table above.
+For the VLM-AUC evaluation, each scenario additionally contributes **72 human-labeled samples** (binary collision / no-collision); the positive/negative ratio reflects the natural contact frequency and is not balanced.
 
 ### Loading the Data
 
-This is a large, Git-LFS–backed dataset — **avoid cloning or downloading everything**. Fetch only the specific archives you need with `huggingface_hub`:
+This dataset is large and Git-LFS–backed — **avoid cloning or downloading everything**. Fetch only the specific archives you need with `huggingface_hub`:
 
 ```python
 from huggingface_hub import hf_hub_download
@@ -259,58 +300,46 @@ files = HfApi().list_repo_files("zzhou292/DreamerBench", repo_type="dataset")
 print("\n".join(files))
 ```
 
-## Dataset Creation
-
-### Curation Rationale
+## Associated Model and Benchmark
 
-DreamerBench was created to provide **physics-grounded** training data for world models. Most video and interaction datasets capture appearance but omit the underlying contact and friction dynamics that govern manipulation. By generating trajectories in a high-fidelity physics simulator, DreamerBench pairs visual observations with ground-truth contact forces, friction coefficients, and slip — enabling models to learn dynamics rather than only visual correlations.
+### ChronoDreamer World Model
 
-### Source Data
+DreamerBench is the training and evaluation substrate for **ChronoDreamer**, an action-conditioned world model. Given a history of egocentric RGB frames, actions, and joint states, ChronoDreamer jointly predicts **future video tokens, contact-splat tokens, and joint angles**. It operates on the precomputed Cosmos DI8×8 tokens with a **spatial-temporal transformer** (24 layers, hidden size 256, 8 heads; factorized spatial and causal-temporal attention) trained with a **MaskGIT-style** masked-prediction objective over a factorized token vocabulary. The model has roughly **30–50M parameters** and was trained for 10 epochs on four NVIDIA A100 GPUs (~10 days) over ~12 hours of interaction data.
 
-All data is **synthetically generated** using [Project Chrono](https://projectchrono.org/), an open-source multi-physics simulation engine. Scenarios are named after the interacting objects (`flashlight_box`, `flashlight_coca`, `waterbottle_coca`, and `fea_flashlight` — a flashlight variant used for the ablation study documented in `pretrained_ckpt/fea_flashlight/`). Each episode is rendered from multiple synchronized viewpoints (Ego, Side 1, Side 2) together with a contact-splat visualization, and latent encodings are produced with the Cosmos Tokenizer DI 8×8 autoencoder. No human-subject, web-scraped, or otherwise externally sourced data is involved.
+Pretrained checkpoints are provided under `pretrained_ckpt/` (`mode_0`, `mode_1`, and an `fea_flashlight` ablation).
 
-### Personal and Sensitive Information
+### VLM-AUC Evaluation
 
-The dataset contains **no personal, private, or sensitive information**. All content is procedurally generated in simulation and does not depict real people, locations, or identifiable entities.
+**VLM-AUC** is an offline protocol for measuring world-model quality without ground-truth forces at test time. A predicted rollout is scored for collision severity on a symmetric 1–5 scale by an **ensemble of five prompt variants** (baseline, evidence-first CoT, conservative gating, future-only temporal boundary, and counterfactual verification); the averaged ensemble score's **AUC** against human binary collision labels is computed via the **Wilcoxon–Mann–Whitney** statistic, making it robust to class imbalance.
 
-## Considerations for Using the Data
+Three open-weight judges are evaluated — **Llama 3.2 90B**, **Gemma 3 27B**, and **Qwen3.5 122B** — across three conditions: ground truth (GT), epoch-10 predictions, and epoch-1 predictions. The protocol expects the ordering `AUC(GT) ≥ AUC(Ep.10) ≥ AUC(Ep.1)`. At the aggregate level this ordering holds for the larger judges, while the smallest judge inverts the epoch sub-ordering, consistent with reduced calibration at smaller scale. Complete evaluation artifacts — inference logs, reasoning traces, raw outputs, and parsed scores — are archived in `inference_results.zip`.
 
-### Social Impact of Dataset
+## Intended Uses and Tasks
 
-DreamerBench is intended to advance research in physically grounded world models, model-based reinforcement learning, and robotic manipulation. Because it is fully synthetic, it carries low direct social risk and can reduce the cost and safety concerns associated with large-scale real-robot data collection.
+DreamerBench supports world models that must account for both visual appearance and low-level contact physics:
 
-### Discussion of Biases
+- **One-step and multi-step prediction** in pixel or latent space (image reconstruction error, proprioception error, contact-mode cross-entropy; rollout-degradation metrics such as SSIM and force-sequence deviation).
+- **Contact- and friction-specific evaluation**: contact-onset and mode prediction (precision/recall/F1 on `c`), normal/tangential force regression, and robustness under friction-domain shift.
+- **Contact-implicit / contact-aware planning** and **model-based reinforcement learning** from offline trajectories.
+- **World-model assessment** via the VLM-AUC protocol.
 
-The data reflects the modeling assumptions of the simulator (material properties, contact models, friction parameters, camera placement, and the specific object set). Models trained solely on this data may inherit those assumptions and may not generalize to dynamics or objects outside the simulated distribution.
+## Limitations
 
-### Other Known Limitations
-
-- **Sim-to-real gap:** simulated contact/friction dynamics are approximations of real-world physics.
-- **Limited diversity:** a small set of curated interaction scenarios and objects.
-- **Low observation resolution:** stored observations are 64×64 (previews are rendered at 256×256).
-- **Storage/bandwidth:** files are distributed as Git-LFS ZIP archives; download selectively (see [Loading the Data](#loading-the-data)).
+- **Simulation only.** All data is generated in Project Chrono and inherits its modeling assumptions; a sim-to-real gap remains.
+- **Modest volume and contact sparsity.** ~12 hours of data, and contact events are a small fraction of frames — likely contributing to post-contact blurring observed in trained models.
+- **Narrow coverage.** Four scenarios and a single robot morphology; far fewer than broad manipulation benchmarks, and generalization across embodiments is not probed.
+- **Evaluation breadth.** The released results focus on video cross-entropy, LPIPS, joint-angle MSE, and VLM-AUC; the full contact/friction metric suite and external world-model baselines (Dreamer, PlaNet, Genie-style) are left to future work.
 
 ## Additional Information
 
-### Dataset Curators
-
-DreamerBench is developed as part of the **ChronoDreamer** project (https://github.com/uwsbel/ChronoDreamer). Point of contact: **Json Zhou** (`zzhou292@wisc.edu`).
-
-### Licensing Information
+### Licensing
 
 Released under the **MIT License**.
 
-### Citation Information
+### Citation
 
-If you use DreamerBench, please cite the project. A formal paper is forthcoming; until then, the placeholder below may be used (update the title, authors, and year once published):
+A citation for DreamerBench will be provided here at a later date.
 
-```bibtex
-@misc{dreamerbench2025,
-  title        = {DreamerBench: A Physics-Grounded Dataset for World Models},
-  author       = {Zhou, Json},
-  howpublished = {Hugging Face Datasets},
-  year         = {2025},
-  url          = {https://huggingface.co/datasets/zzhou292/DreamerBench},
-  note         = {Project: https://github.com/uwsbel/ChronoDreamer}
-}
-```
+### Acknowledgments
+
+Developed by the **Simulation-Based Engineering Laboratory (SBEL)** at the University of Wisconsin–Madison. This work was supported in part by the National Science Foundation under Award No. CMMI-2153855, and used the Euler cluster at UW–Madison and the [Project Chrono](https://projectchrono.org/) simulation infrastructure.