Improve model card for Variational-Reasoning-4B-Acc

README.md

@@ -1,199 +1,98 @@
 ---
 library_name: transformers
-tags: []
+pipeline_tag: text-generation
+license: apache-2.0
+base_model: Qwen3-4B-Base
+tags:
+- qwen3
+- reasoning
+- variational-inference
 ---
 
-# Model Card for Model ID
-
-<!-- Provide a quick summary of what the model is/does. -->
-
+# Model Card for Variational-Reasoning-4B-Acc
 
+This model is `Variational-Reasoning-4B-Acc`, a part of the "Variational Reasoning for Language Models" framework, which was introduced in the paper [Variational Reasoning for Language Models](https://huggingface.co/papers/2509.22637). This framework treats thinking traces as latent variables and optimizes them through variational inference, aiming to provide stable objectives for improving the reasoning ability of language models.
 
 ## Model Details
 
 ### Model Description
 
-<!-- Provide a longer summary of what this model is. -->
-
-This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
-
-- **Developed by:** [More Information Needed]
-- **Funded by [optional]:** [More Information Needed]
-- **Shared by [optional]:** [More Information Needed]
-- **Model type:** [More Information Needed]
-- **Language(s) (NLP):** [More Information Needed]
-- **License:** [More Information Needed]
-- **Finetuned from model [optional]:** [More Information Needed]
+The `Variational-Reasoning-4B-Acc` model implements a variational reasoning framework that extends the evidence lower bound (ELBO) to a multi-trace objective and proposes a forward-KL formulation to stabilize the training of the variational posterior. It provides a principled probabilistic perspective that unifies variational inference with RL-style methods. This specific model checkpoint is based on the `Qwen3-4B-Base` backbone and is optimized using an accuracy-based estimator. It has been empirically validated across a wide range of reasoning tasks.
 
-### Model Sources [optional]
+-   **Developed by:** Xiangxin Zhou, Zichen Liu, Haonan Wang, Chao Du, Min Lin, Chongxuan Li, Liang Wang, Tianyu Pang
+-   **Model type:** Causal Language Model (Qwen2ForCausalLM)
+-   **Language(s) (NLP):** English
+-   **License:** Apache 2.0
+-   **Finetuned from model:** Qwen3-4B-Base
 
-<!-- Provide the basic links for the model. -->
+### Model Sources
 
-- **Repository:** [More Information Needed]
-- **Paper [optional]:** [More Information Needed]
-- **Demo [optional]:** [More Information Needed]
+-   **Repository:** [https://github.com/sail-sg/variational-reasoning](https://github.com/sail-sg/variational-reasoning)
+-   **Paper:** [https://huggingface.co/papers/2509.22637](https://huggingface.co/papers/2509.22637)
 
 ## Uses
 
-<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
-
 ### Direct Use
 
-<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
-
-[More Information Needed]
-
-### Downstream Use [optional]
-
-<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
-
-[More Information Needed]
+This model is primarily intended for research and development in improving the reasoning capabilities of language models. It can be directly used for tasks requiring advanced reasoning, multi-step problem solving, and generating coherent "thinking traces".
 
 ### Out-of-Scope Use
 
-<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
-
-[More Information Needed]
+The model is not intended for generating harmful content, hate speech, or discriminatory language. Users should exercise caution and ensure responsible deployment. It is not designed for real-world factual retrieval without external knowledge sources or applications requiring perfect accuracy in highly sensitive domains.
 
 ## Bias, Risks, and Limitations
 
-<!-- This section is meant to convey both technical and sociotechnical limitations. -->
-
-[More Information Needed]
+As with all large language models, this model may inherit biases present in its training data. Users should be aware of potential biases and limitations in its reasoning abilities, especially in complex or sensitive contexts. Further analysis of its behavior in specific application areas is recommended.
 
 ### Recommendations
 
-<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
-
-Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+Users (both direct and downstream) should be made aware of the risks, biases, and limitations of the model. Continuous monitoring and evaluation for unintended consequences are recommended.
 
 ## How to Get Started with the Model
 
-Use the code below to get started with the model.
-
-[More Information Needed]
+For detailed instructions on how to load, use, train, and evaluate the model, please refer to the official [GitHub repository](https://github.com/sail-sg/variational-reasoning). The repository provides comprehensive scripts and guidelines to reproduce the experiments and work with the model effectively.
 
 ## Training Details
 
 ### Training Data
 
-<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
-
-[More Information Needed]
+The `Variational-Reasoning-4B-Acc` model was trained using specific datasets, including `Variational-Posterior-4B-Acc-mix` (as indicated in the GitHub repository's model table). The training process involves multiple stages that generate and utilize intermediate datasets for the initial reasoning model and the variational posterior. Detailed information about data processing and dataset specifics can be found in the [GitHub repository](https://github.com/sail-sg/variational-reasoning).
 
 ### Training Procedure
 
-<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
-
-#### Preprocessing [optional]
-
-[More Information Needed]
+The training procedure for the variational reasoning framework is multi-staged and builds upon `LLaMA-Factory` and `SkyThought`. It involves:
+1.  Training an initial reasoning model ($\pi_{\theta_0}$).
+2.  Training a variational posterior ($q_\phi$).
+3.  Sampling from the variational posterior.
+4.  Estimating log likelihoods using both models.
+5.  (Optional) Sampling from the initial reasoning model and verification.
+6.  Building the dataset for training the final reasoning model ($\pi_\theta$) using an accuracy-based estimator (for this specific model).
+7.  Training the final reasoning model ($\pi_\theta$).
 
+Detailed scripts, configurations (e.g., `LLaMA-Factory/examples/variational_reasoning/*.yaml`), and instructions for each step are available in the [GitHub repository's training section](https://github.com/sail-sg/variational-reasoning#training).
 
 #### Training Hyperparameters
 
-- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
-
-#### Speeds, Sizes, Times [optional]
-
-<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
-
-[More Information Needed]
+Training scripts assume 2 nodes with 8 H100 GPUs each. Users with different hardware configurations are advised to adjust `gradient_accumulation_steps` to maintain a constant effective batch size. Specific hyperparameters are detailed in the YAML configuration files within the GitHub repository.
 
 ## Evaluation
 
-<!-- This section describes the evaluation protocols and provides the results. -->
-
 ### Testing Data, Factors & Metrics
 
-#### Testing Data
-
-<!-- This should link to a Dataset Card if possible. -->
-
-[More Information Needed]
-
-#### Factors
-
-<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
-
-[More Information Needed]
-
-#### Metrics
-
-<!-- These are the evaluation metrics being used, ideally with a description of why. -->
-
-[More Information Needed]
+The models are empirically validated on the Qwen 2.5 and Qwen 3 model families across a wide range of reasoning tasks. The evaluation process is conducted using the SkyThought evaluation suite. Specific evaluation protocols, testing data, factors, and metrics (e.g., accuracy based on verification) are detailed in the [GitHub repository's evaluation section](https://github.com/sail-sg/variational-reasoning#evaluation), particularly referring to `SkyThought/variational_reasoning/eval/eval.sh`.
 
 ### Results
 
-[More Information Needed]
-
-#### Summary
-
-
-
-## Model Examination [optional]
-
-<!-- Relevant interpretability work for the model goes here -->
-
-[More Information Needed]
-
-## Environmental Impact
-
-<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
-
-Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
-
-- **Hardware Type:** [More Information Needed]
-- **Hours used:** [More Information Needed]
-- **Cloud Provider:** [More Information Needed]
-- **Compute Region:** [More Information Needed]
-- **Carbon Emitted:** [More Information Needed]
-
-## Technical Specifications [optional]
-
-### Model Architecture and Objective
-
-[More Information Needed]
-
-### Compute Infrastructure
-
-[More Information Needed]
-
-#### Hardware
-
-[More Information Needed]
-
-#### Software
-
-[More Information Needed]
-
-## Citation [optional]
-
-<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
-
-**BibTeX:**
-
-[More Information Needed]
-
-**APA:**
-
-[More Information Needed]
-
-## Glossary [optional]
-
-<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
-
-[More Information Needed]
-
-## More Information [optional]
-
-[More Information Needed]
-
-## Model Card Authors [optional]
-
-[More Information Needed]
+The paper and the GitHub repository provide empirical validation and results demonstrating improved reasoning abilities. For specific performance metrics and comparisons, refer to the original paper and the GitHub repository.
 
-## Model Card Contact
+## Citation
 
-[More Information Needed]
+If you find this work useful, please consider citing our paper:
+```bib
+@article{zhou2025variationalreasoninglanguagemodels,
+      title={Variational Reasoning for Language Models}, 
+      author={Xiangxin Zhou and Zichen Liu and Haonan Wang and Chao Du and Min Lin and Chongxuan Li and Liang Wang and Tianyu Pang},
+      journal={arXiv preprint arXiv:2509.22637},
+      year={2025}
+}
+```