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..
Neural Belief Reasoner
Authors: Haifeng Qian
IJCAI 2020 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | This paper studies NBR in two tasks. The first is a synthetic unsupervisedlearning task, which demonstrates NBR s ability to perform multi-hop reasoning, reasoning with uncertainty and reasoning about conflicting information. The second is supervised learning: a robust MNIST classifier for 4 and 9, which is the most challenging pair of digits. |
| Researcher Affiliation | Industry | Haifeng Qian IBM Research, Yorktown Heights, NY, USA EMAIL |
| Pseudocode | No | No pseudocode or clearly labeled algorithm blocks were found in the paper. |
| Open Source Code | Yes | Source code for training and in-ference is available at http://researcher.watson.ibm.com/group/10228 |
| Open Datasets | Yes | The second task is supervised learning: a robust MNIST classifier for 4 and 9, which is the most challenging pair of digits. |
| Dataset Splits | No | The paper mentions 'training images' and refers to the 'MNIST training set' but does not specify exact training/validation/test splits (e.g., percentages or counts) or refer to a standard split with a citation providing such details. |
| Hardware Specification | No | No specific hardware details (e.g., GPU/CPU models, memory) used for running the experiments were provided. |
| Software Dependencies | No | No specific software dependencies with version numbers (e.g., Python 3.x, PyTorch 1.x, CUDA 11.x) were provided. |
| Experiment Setup | Yes | The first seven Gi ( ) s are trained jointly with the following loss function: ... where s and β are hyperparameters. ... where ω is a hyperparameter. ... Iteration limit is 100 for PGD, 50K for BA, and 10K for CW and SCW. |