Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Fast Monte Carlo Tree Diffusion: 100× Speedup via Parallel and Sparse Planning
Authors: Jaesik Yoon, Hyeonseo Cho, Yoshua Bengio, Sungjin Ahn
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments show that Fast-MCTD achieves up to 100 speedup over standard MCTD while maintaining or improving planning performance. |
| Researcher Affiliation | Collaboration | Jaesik Yoon KAIST & SAP EMAIL |
| Pseudocode | Yes | Finally, we integrate the aforementioned Parallel MCTD (P-MCTD; Section 4.1) and Sparse MCTD (S-MCTD; Section 4.2) approaches into our final proposed method, termed Fast Monte Carlo Tree Diffusion (Fast-MCTD). We illustrate the details of Fast-MCTD in Algorithm 1. ... Algorithm 2 Fast Monte Carlo Tree Diffusion |
| Open Source Code | Yes | The source code is publicly available at https://github.com/ahn-ml/mctd. |
| Open Datasets | Yes | We evaluate Fast-MCTD using tasks from the Offline Goal-conditioned RL benchmark (OGBench) [22], aligning with the setup in MCTD [33]. |
| Dataset Splits | Yes | We evaluate Fast-MCTD using tasks from the Offline Goal-conditioned RL benchmark (OGBench) [22], aligning with the setup in MCTD [33]. ... All evaluation tasks were unseen during training, requiring the models to generalize their planning capabilities to novel scenarios at inference time. |
| Hardware Specification | Yes | The experiments were conducted on a server equipped with 8 NVIDIA RTX 4090 GPUs, 512 GB of system memory, and a 96-thread CPU. |
| Software Dependencies | No | The paper discusses various models and techniques like Diffusion Forcing, DDIM, Transformer-based architecture, and mentions using Python implicitly, but does not provide specific version numbers for any software, libraries, or frameworks used. |
| Experiment Setup | Yes | Table 7: Hyperparameter configurations for the Diffuser. Table 8: Hyperparameters for the value-learning policy. Table 9: Hyperparameters for the Diffusion Forcing. Table 10: Hyperparameters for MCTD and its variants. |