CoolFace
Modelpublic

X1AOX1A/WorldModel-Sciworld-Llama3.1-8B

sourceHugging Faceotherupdated 9mo agoView on Hugging Face
0likes13downloads
Model Card

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

From Word to World: Can Large Language Models be Implicit Text-based World Models?

![arXiv](https://arxiv.org/abs/2512.18832) ![Blog](https://macaron.im/mindlab/research/how-world-models-unlock-scalable-agentic-rl) ![HF Paper](https://huggingface.co/papers/2512.18832) ![Models](https://huggingface.co/collections/X1AOX1A/llm-as-world-models) ![Dataset](https://huggingface.co/datasets/X1AOX1A/LLMasWorldModels)

sciworldtrainwithenv40k

This model is a fine-tuned version of meta-llama/Meta-Llama-3.1-8B on the sciworldtrainwithenv40630 dataset.

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • —learning_rate: 1e-05
  • —trainbatchsize: 2
  • —evalbatchsize: 8
  • —seed: 42
  • —distributed_type: multi-GPU
  • —num_devices: 4
  • —gradientaccumulationsteps: 16
  • —totaltrainbatch_size: 128
  • —totalevalbatch_size: 32
  • —optimizer: Use OptimizerNames.ADAMWTORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizerargs=No additional optimizer arguments
  • —lrschedulertype: constantwithwarmup
  • —lrschedulerwarmup_steps: 10
  • —num_epochs: 5.0

Training results

Framework versions

  • —Transformers 4.52.4
  • —Pytorch 2.9.0+cu128
  • —Datasets 3.6.0
  • —Tokenizers 0.21.1