---
library_name: transformers
license: apache-2.0
tags:
- memory-decoder
- parametric-memory
- long-term-memory
- memorydecoder-at-scale
- text-generation
- olmo
- finance
datasets:
- Rubin-Wei/MemoryDecoder-at-Scale-domain-data
---
# MemoryDecoder-OLMo-1.7B-finance

## Resources

- [Project Page: Memory Decoder at Scale](https://rubin-wei.github.io/memory-decoder-at-scale/)
- [GitHub Repository: LUMIA-Group/MemoryDecoder-at-Scale](https://github.com/LUMIA-Group/MemoryDecoder-at-Scale)
- [Paper: Memory Decoder at Scale: A Pretrained, Parametric Long-Term Memory](https://arxiv.org/abs/2607.27919)
- [Hugging Face Collection: MemoryDecoder-at-Scale](https://huggingface.co/collections/Rubin-Wei/memorydecoder-at-scale)
- [Dataset: MemoryDecoder-at-Scale Domain Data (Finance)](https://huggingface.co/datasets/Rubin-Wei/MemoryDecoder-at-Scale-domain-data/tree/main/finance)

This repository contains the **1.7B finance Memory Decoder** released
with [Memory Decoder at Scale](https://arxiv.org/abs/2607.27919). It is a pretrained parametric long-term
memory that can be swapped into a compatible frozen language-model backbone.

This OLMo-vocabulary memory was trained for two epochs on the finance CPT corpus. The released configuration uses the OLMo vocabulary so that the memory can be combined with a vocabulary-compatible frozen OLMo backbone.

This checkpoint is a memory component, not a standalone chat- or
instruction-tuned model. The memory and backbone must use compatible token IDs
and vocabularies.

## Model details

| Field | Value |
|---|---|
| Memory size | 1.7B class |
| Architecture/tokenizer family | Qwen3-style 1.7B memory; OLMo tokenizer/vocabulary |
| Domain | Finance |
| Evaluation benchmark | FinEval |
| Intended backbone | Frozen OLMo-family model with the matching tokenizer/vocabulary |
| Release contents | Inference weights, configuration, and tokenizer files |

## Usage

Install the matching environment from
[LUMIA-Group/MemoryDecoder-at-Scale](https://github.com/LUMIA-Group/MemoryDecoder-at-Scale), then
set `MODEL_PATH` to a compatible frozen backbone and `MEMDEC_PATH` to this
repository:

```bash
MODEL_PATH=/path/to/compatible-base-model \
MEMDEC_PATH=Rubin-Wei/MemoryDecoder-OLMo-1.7B-finance \
bash eval/lm-evaluation-harness/scripts/domain/evaluate_fineval.sh
```

See the repository documentation and launcher for benchmark-specific options,
including interpolation weights and batch settings.

## Intended use and limitations

This checkpoint is intended for research and evaluation in the finance
domain. Its outputs depend on the backbone, prompt, and interpolation settings.
Domain specialization does not guarantee factual correctness or safety, and the
model may inherit biases and errors from its training sources.

## Citation

If you use this checkpoint, please cite:

```bibtex
@misc{wei2026memorydecoderscalepretrained,
      title={Memory Decoder at Scale: A Pretrained, Parametric Long-Term Memory},
      author={Rubin Wei and Jiaqi Cao and Jiarui Wang and Junming Zhang and Qipeng Guo and Bowen Zhou and Zhouhan Lin},
      year={2026},
      eprint={2607.27919},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2607.27919},
}
```