MOOSE-Star-HC-R1D-7B Model Card

Overview

MOOSE-Star-HC-R1D-7B (referred to as MS-HC-7B w/ 1x bounded in the paper) is a 7B parameter language model fine-tuned for generating scientific hypotheses from research questions, background surveys, and cross-paper inspirations. It's designed for incremental hypothesis composition in scientific discovery workflows.

• Paper: MOOSE-Star: Unlocking Tractable Training for Scientific Discovery by Breaking the Complexity Barrier (arXiv:2603.03756)
• Base Model: DeepSeek-R1-Distill-Qwen-7B
• License: Apache 2.0
• Code: ZonglinY/MOOSE-Star
• Multi-task variant: MOOSE-Star-R1D-7B (IR + HC in one model)

Model Description

Parameter	Value
Base Model	DeepSeek-R1-Distill-Qwen-7B
Training Method	Full-parameter SFT (ZeRO-3)
Training Data	TOMATO-Star-SFT-Data-R1D-32B HC split (114,548 samples, with 1x bounded composition)
Chat Template	deepseekr1
Cutoff Length	8192
Learning Rate	1e-5
Epochs	1
Batch Size	128
Training	Full-parameter, ZeRO-3, bf16

Task Description

The model generates delta hypotheses — specific contributions from individual inspiration papers. Given:

• Research question
• Background survey
• Previous hypothesis (optional)
• New inspiration paper (title + abstract)

The model outputs:

• Inspiration: Key concept from the paper
• Motivation (WHY): How it addresses gaps
• Mechanism (HOW IT WORKS): Core scientific mechanism
• Methodology (HOW IT'S INTEGRATED): Implementation approach

Prompt Format (Simplified Overview)

The full prompt template is constructed via instruction_prompts() in the code example below. The general structure is:

[Task instruction preamble]

## Information Provided

**Research Question**:
{research_question}

**Background Survey**:
{background_survey}

**Previous Hypothesis**:
{previous_hypothesis_or_none}

**New Inspiration Paper Title**:
{inspiration_title}

**New Inspiration Paper Abstract**:
{inspiration_abstract}

## Your Response

<think>
[reasoning process]
</think>

Inspiration: [Key concept]
- Motivation (WHY): [Why this addresses a gap]
- Mechanism (HOW IT WORKS): [How the concept works]
- Methodology (HOW IT'S INTEGRATED): [Implementation steps]

Usage

Prerequisites: Clone the MOOSE-Star repo for prompt templates:

git clone https://github.com/ZonglinY/MOOSE-Star.git && cd MOOSE-Star
# See requirements.txt for full dependencies; at minimum: pip install transformers torch

import sys
sys.path.insert(0, "./utils")
from prompt_store import instruction_prompts
from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "ZonglinY/MOOSE-Star-HC-R1D-7B"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name, 
    dtype="auto", 
    device_map="auto"
)

# Use bounded/delta composition for hierarchical search
p = instruction_prompts("prepare_HC_sft_data_to_go_comprehensive_v2_delta")

research_question = "Your research question here"
background_survey = "Your background survey here"
inspiration_title = "Inspiration paper title"
inspiration_abstract = "Inspiration paper abstract"

prompt = (p[0] + research_question
        + p[1] + background_survey
        + p[2] + "No previous hypothesis."
        + p[3] + inspiration_title
        + p[4] + inspiration_abstract
        + p[5])

messages = [{"role": "user", "content": prompt}]
formatted = tokenizer.apply_chat_template(
    messages, 
    tokenize=False, 
    add_generation_prompt=False
)
formatted += "<｜Assistant｜>"

inputs = tokenizer(formatted, return_tensors="pt").to(model.device)
outputs = model.generate(
    **inputs, 
    max_new_tokens=8192, 
    temperature=0.6, 
    top_p=0.9, 
    do_sample=True
)
response = tokenizer.decode(
    outputs[0][inputs["input_ids"].shape[1]:], 
    skip_special_tokens=True
)
print(response)

Evaluation Results

Rubric-based evaluation on normal composition with ground-truth inspirations (Judge: GPT-4o):

Model	Total	Mot	Mec	Met	Length
R1-Distilled-Qwen-7B (base)	4.05	1.96	1.30	0.80	231.02
MS-HC-7B	4.68	2.13	1.46	1.09	204.12
MS-HC-7B w/ 1x bounded (this model)	4.74	2.16	1.48	1.10	203.84
MS-HC-7B w/ 2x bounded	4.73	2.15	1.48	1.09	205.17

Citation

@article{yang2025moosestar,
  title={MOOSE-Star: Unlocking Tractable Training for Scientific Discovery by Breaking the Complexity Barrier},
  author={Yang, Zonglin and Bing, Lidong},
  journal={arXiv preprint arXiv:2603.03756},
  year={2026}
}