MemSFT specializes modern large language models with an external parametric memory. This checkpoint contains the Qwen3-8B memory trained on Biology-Instructions. The memory learns to approximate retrieval-based teacher distributions over domain SFT data. At each decoding step, a learned token-level router combines the next-token distributions of the frozen base model and memory. This checkpoint is an auxiliary memory, not a standalone chat model. Its key advantages are:

Plug-and-Play: Attaches to a frozen backbone without modifying its parameters or architecture.
Strong Specialization: Improves domain performance with negligible degradation in general capabilities.
Cross-Scale Reuse: Works with Qwen3 backbones from 8B to 235B-A22B without retraining the memory for each backbone.

Quick Start

The 14B + 8B example is intended for a CUDA GPU with sufficient memory to load both models in BF16.

1. Install

bash

git clone https://github.com/LUMIA-Group/MemSFT.git
cd MemSFT

conda create -n memsft-generate python=3.10 pip -y
conda activate memsft-generate
python -m pip install -e .
python -m pip install \
  "torch>=2.4,<2.7" \
  "transformers==4.51.3" \
  "huggingface-hub==0.35.3" \
  "accelerate>=0.34,<2"

2. Load the base, memory, and router

python

from pathlib import Path

import torch
from huggingface_hub import snapshot_download
from transformers import AutoModelForCausalLM, AutoTokenizer

from memsft.router.adaptive_memdec import AdaptiveMemoryDecoder

device = torch.device("cuda:0")
base_id = "Qwen/Qwen3-14B"
memory_id = "Jiarui-Wang/MemSFT-Qwen3-Bio-Memory-8B"
router_repo = "Jiarui-Wang/MemSFT-Qwen3-Routers"
router_subdir = "Qwen3-14B-Bio-M8B-Router"

router_root = snapshot_download(
    repo_id=router_repo,
    revision="v1.0.0",
    allow_patterns=[f"{router_subdir}/*"],
)
router_path = str(Path(router_root) / router_subdir)

tokenizer = AutoTokenizer.from_pretrained(
    base_id,
    revision="40c069824f4251a91eefaf281ebe4c544efd3e18",
)
base = AutoModelForCausalLM.from_pretrained(
    base_id,
    revision="40c069824f4251a91eefaf281ebe4c544efd3e18",
    torch_dtype=torch.bfloat16,
    low_cpu_mem_usage=True,
).to(device).eval()
memory = AutoModelForCausalLM.from_pretrained(
    memory_id,
    revision="v1.0.0",
    torch_dtype=torch.bfloat16,
    low_cpu_mem_usage=True,
).to(device).eval()

vocab_size = len(tokenizer)
base.resize_token_embeddings(vocab_size)
memory.resize_token_embeddings(vocab_size)
base.requires_grad_(False)
memory.requires_grad_(False)
model = AdaptiveMemoryDecoder(
    base_lm=base,
    knn_generator=memory,
    router_path=router_path,
    router_device=device,
).eval()
model.set_tokenizer(tokenizer)

3. Generate

python

sequence = (
    "MKSILIEKPNQLAIVEREIPTPSAGEVRVKVKLAGICGSDSHIYRGHNPFAKYPRVIGHEFFGVIDAV"
    "GEGVESARVGERVAVDPVVSCGHCYPCSIGKPNVCTTLAVLGVHADGGFSEYAVVPAKNAWKIPEAVA"
    "DQYAVMIEPFTIAANVTGHGQPTENDTVLVYGAGPIGLTIVQVLKGVYNVKNVIVADRIDERLEKAKE"
    "SGADWAINNSQTPLGEIFTEKGIKPTLIIDAACHPSILKEAVTLASPAARIVLMGFSSEPSEVIQQGI"
    "TGKELSIFSSRLNANKFPIVIDWLSKGLIKPEKLITHTFDFQHVADAISLFEQDQKHCCKVLLTFSE"
)
prompt = (
    r"<PROTEIN> "
    + sequence
    + r" </PROTEIN> What is the EC number associated with the enzymatic "
    r"function of this protein? Please put the final enzyme within \boxed{} "
    r"using an EC number such as ECx.x.x.x, and separate multiple entries "
    r"with commas."
)

messages = [{"role": "user", "content": prompt}]
prompt_text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
    enable_thinking=False,
)
inputs = tokenizer(prompt_text, return_tensors="pt").to(device)

with torch.inference_mode():
    output_ids = model.generate(
        **inputs,
        do_sample=False,
        max_new_tokens=32,
        eos_token_id=tokenizer.eos_token_id,
        pad_token_id=tokenizer.eos_token_id,
    )

answer = tokenizer.decode(
    output_ids[0, inputs["input_ids"].shape[1]:],
    skip_special_tokens=True,
)
print(answer)

Example output:

\boxed{EC1.1.1.-}

UniProtKB annotates Escherichia coli K-12 RspB (P38105) with EC 1.1.1.-. For comparison, using the same prompt and deterministic generation configuration, Qwen3-14B alone predicts EC 4.2.1.22. The outputs were reproduced in BF16 on NVIDIA A800 80GB GPUs.

Performance

The same Qwen3-8B Biology-Instructions memory is reused with each base model; each pairing uses its corresponding router.

Base model	Biology-Instructions ↑	General ↑
Qwen3-8B + MemSFT	42.84	81.15
Qwen3-14B + MemSFT	42.92	83.62
Qwen3-32B + MemSFT	42.82	85.33
Qwen3-235B-A22B + MemSFT	42.05	87.11

Compatible Pairing

The example above uses:

base: Qwen/Qwen3-14B
memory: Jiarui-Wang/MemSFT-Qwen3-Bio-Memory-8B
router: Jiarui-Wang/MemSFT-Qwen3-Routers/Qwen3-14B-Bio-M8B-Router

MemSFT prefers tensor-only .safetensors router checkpoints. Legacy .pt checkpoints should be loaded only from trusted sources; the MemSFT loader uses PyTorch's restricted weights_only=True mode for compatibility.

Intended Use and Limitations

This checkpoint is intended for reproducing MemSFT and for augmenting compatible Qwen3 base models on Biology-Instructions tasks. It should be used with the router matching the selected base/memory pair. Performance outside the evaluated model combinations and domains has not been established.

License

This MemSFT checkpoint is released under the Apache License 2.0. Upstream models, software, and datasets remain subject to their respective licenses and terms.

Citation

If you find MemSFT helpful in your research, please consider citing:

bibtex

@misc{wang2026memsftmitigatingalignmenttax,
      title={MemSFT: Mitigating Alignment Tax with an External Parametric Memory},
      author={Jiarui Wang and Xiang Shi and Jiaqi Cao and Rubin Wei and Xiquan Wang and Hao Sun and Jingzhi Wang and Zhiqi Yang and Qipeng Guo and Bowen Zhou and Zhouhan Lin},
      year={2026},
      eprint={2607.25614},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2607.25614}, 
}

Contact

For questions and discussions, feel free to email wangjiarui1@sjtu.edu.cn.