ExoMind: Democratizing Scientific Intelligence via Extended-Mind-Inspired Agentic System

Name: ExoMind
Availability: InStock

ExoMind Team · Shanghai Artificial Intelligence Laboratory

Project Page

ExoMind preprint

Hugging Face

GitHub

🔥 News

2026-08-28: 🔥 The ExoMind preprint is now available on Preprints.org.
2026.8.25: 🔥 We have released a series of official GGUF variants for ExoMind and ExoMind-9B. Please refer to the ExoMind collection. We also thank mradermacher for providing additional GGUF quantizations, including both static and importance-matrix variants: ExoMind-i1-GGUF, ExoMind-GGUF, ExoMind-9B-i1-GGUF, and ExoMind-9B-GGUF.
2026-08-12: We release the official project page and public repository.

Overview

ExoMind is the first extended-mind-inspired agentic system designed for scientific reasoning and research. It organizes a general-purpose model, specialized interaction objects, and autonomous interaction processes as one system, allowing the model to discover sources, ground evidence, execute verification, and update its reasoning around each scientific problem.

This repository hosts the main checkpoint, fine-tuned from Qwen3.5-35B-A3B. With training-value-aware data engineering, a scientific interaction framework, and two-stage progressive Chain-of-Interaction training, ExoMind raises the average score across eight scientific benchmarks from 36.2 to 68.3, achieves the highest average among all evaluated models, and ranks first on six benchmarks.

Highlights

Extended-mind-inspired intelligence: unifies the LLM, interaction objects, and autonomous interaction processes as a scientific agentic system.
Training-value-aware data engineering: identifies challenging, learnable problems and routes them to pure-reasoning or interaction-reasoning data.
Scientific interaction: turns source discovery, evidence grounding, executable verification, and observation integration into composable objects.
Progressive CoI training: jointly develops intrinsic reasoning and autonomous interaction using a few thousand high-quality trajectories.
Efficient frontier performance: completes two-stage full-parameter SFT in 1–2 days on 8 NVIDIA H200 GPUs while improving all six evaluated general capability benchmarks over the base model.

ExoMind scientific intelligence evaluation

Quick Start

Use a recent vLLM or SGLang release with Qwen3.5 support. The examples below use the checkpoint's configured maximum context length of 262,144 tokens.

vLLM

bash

vllm serve AI4SGI/ExoMind \
  --port 8000 \
  --tensor-parallel-size 8 \
  --max-model-len 262144 \
  --reasoning-parser qwen3 \
  --enable-auto-tool-choice \
  --tool-call-parser qwen3_coder

SGLang

bash

python -m sglang.launch_server \
  --model-path AI4SGI/ExoMind \
  --host 0.0.0.0 \
  --port 8000 \
  --tp-size 8 \
  --context-length 262144 \
  --reasoning-parser qwen3 \
  --tool-call-parser qwen3_coder

OpenAI-Compatible API

python

from openai import OpenAI

client = OpenAI(base_url="http://localhost:8000/v1", api_key="EMPTY")
response = client.chat.completions.create(
    model="AI4SGI/ExoMind",
    messages=[
        {
            "role": "user",
            "content": "Develop and verify a rigorous solution to this scientific problem: ...",
        }
    ],
    temperature=1.0,
    top_p=0.95,
    extra_body={"top_k": 20},
)
print(response.choices[0].message.content)

The complete scientific interaction workflow and prompt contracts are available in the ExoMind GitHub repository.

Evaluation

Under the preprint's evaluation setup, ExoMind reaches an eight-benchmark average of 68.3, compared with 54.2 for the next-best representative model shown below.

🥇 Best score among the representative models shown

Benchmark	⭐ Ours	Representative frontier models
ExoMind 35B-A3B	Claude-Opus-4.8 Thinking	GPT-5.5 (xhigh)	Gemini-3.1-Pro Preview	Kimi-K3	Qwen3.7-Max	GLM-5.2	DeepSeek-V4-Pro (Max)
🧪 Scientific Research
HLE w/ tools	56.8	🥇 57.9	52.2	51.4	56.0	53.5	54.7	48.2
FrontierScience-Research	🥇 70.0	26.7	26.7	11.7	21.7	10.0	15.0	13.3
CMT-Benchmark	🥇 84.0	46.0	43.0	43.0	34.0	34.0	20.0	28.0
CritPt	25.7	20.9	🥇 27.1	17.7	23.4	13.4	20.9	7.1
🧠 Scientific Reasoning
AMO-Bench	🥇 78.0	74.0	70.0	63.1	64.0	57.4	54.0	68.0
IMO-AnswerBench	🥇 92.8	86.8	83.8	90.0	82.8	90.0	91.0	89.8
HiPhO	🥇 49.7	46.4	43.3	43.4	42.4	38.8	37.4	38.7
FrontierScience-Olympiad	🥇 89.0	75.0	78.0	77.0	69.0	80.0	76.5	76.0
Eight-benchmark average	🥇 68.3	54.2	53.0	49.7	49.2	47.1	46.2	46.1

See the interactive evaluation explorer for the complete model list, benchmark scopes, settings, and rankings.

Intended Use

ExoMind is intended for research and development in scientific question answering, literature-grounded investigation, mathematical and computational reasoning, code-assisted verification, and agentic scientific workflows.

License and Attribution

The distributed checkpoint and upstream Qwen3.5 materials are provided under the Apache License 2.0 included in this repository. The preprint, scientific figures and results, and ExoMind brand assets are subject to the ExoMind Research Content and Brand Terms. See NOTICE.md for third-party notices.

Citation

bibtex

@article{Ye_2026,
  title     = {ExoMind: Democratizing Scientific Intelligence via Extended-Mind-Inspired Agentic System},
  author    = {Ye, Peng and Liu, Zhuo and Ye, Jingqi and Yu, Fangchen and Tang, Shengji and Jiang, Yichen and He, Haonan and Cao, Zongsheng and Chen, Tao and Zhang, Bo and Ouyang, Wanli and Zhou, Bowen and Bai, Lei},
  year      = {2026},
  month     = aug,
  publisher = {MDPI AG},
  doi       = {10.20944/preprints202608.2038.v1},
  url       = {https://doi.org/10.20944/preprints202608.2038.v1}
}