15-candidate inspiration retrieval
According to the model card, the model selects one correct cross-paper inspiration from 15 candidates given a research background.
Open Source Model Profile · ZonglinY
MOOSE-Star-IR-R1D-7B
MOOSE-Star-IR-R1D-7B is a 7.62B-parameter Qwen2-family text-generation fine-tune from ZonglinY. According to the model card, it selects one cross-paper inspiration from 15 candidates to support scientific hypothesis generation.
- Publisher
- ZonglinY
- Task
- text-generation
- Model type
- qwen2
- License
- apache-2.0
- Library
- transformers
- Publication status
- Approved for indexing
Model overview
MOOSE-Star-IR-R1D-7B is published by ZonglinY as a Qwen2-family text-generation fine-tune. The captured configuration identifies Qwen2ForCausalLM, and Safetensors metadata reports about 7.62B parameters. According to the model card, it is based on DeepSeek-R1-Distill-Qwen-7B and fine-tuned for cross-paper inspiration retrieval supporting the MOOSE-Star scientific-discovery framework described in arXiv:2603.03756.
Recorded capabilities
Reported 54.37% retrieval accuracy
According to the model card, the model reaches 54.37% accuracy on the inspiration-retrieval evaluation, above the 28.42% reported for its base model.
Full-parameter SFT training
According to the model card, the model was built with full-parameter SFT using ZeRO-3 on DeepSeek-R1-Distill-Qwen-7B.
Documented inference recipes
According to the model card, direct inference uses a chat template with sampling settings such as temperature 0.6 and top_p 0.9, plus an API route via Featherless AI.
Use cases in the source record
- Inspiration-retrieval experiments that select one supporting cross-paper inspiration from 15 candidates given a research background.
- Scientific hypothesis-generation workflows of the kind illustrated by the card's network-neuroscience candidate example.
- Reproducing the publisher's reported retrieval evaluation using the documented SGLang or direct Hugging Face inference recipes.
Limitations and unknowns
- No context-window value was extracted from this record.
- Evaluation figures are publisher-reported claims and were not independently verified.
- Provider state is historical snapshot data, not independently refreshed current availability.
- Training and dataset details come from the publisher model card and were not independently verified.
Source and provenance
Source: ZonglinY/MOOSE-Star-IR-R1D-7B
Captured: Unknown. Processed: 2026-09-07T19:35:53.220075+00:00.
MOOSE-Star-IR-R1D-7B Model Card Overview MOOSE-Star-IR-R1D-7B (referred to as MS-IR-7B in the paper) is a 7B parameter model fine-tuned for selecting the correct cross-paper inspiration from 15 candidates given a research background. It's designed for scientific hypothesis generation in the MOOSE-Star framework. 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 De…
F001F002F003F004F005F006F007F010F011F012F014F016F019F020F021F022F023