Model Card for ethicalabs/Echo-DSRN-v0.1.3-Research-Intent-CLF
This is a ~98 million parameters sequence classification model based on the Echo-DSRN architecture, trained on a single AMD GPU using ROCm 7.2.
OpenAIRE AI Hackathon 2026 🇪🇺
This model is part of ethicalabs.ai's entry in the OpenAIRE AI Hackathon 2026, co-organised by OpenAIRE and Alien Intelligence — a 12-week open science build challenge.
🔴 Live Demo → openaire-2026.ethicalabs.ai
The live application classifies research papers in real time using this model, with publication metadata streamed from the OpenAIRE Graph API under CC BY 4.0.
What this model does
Echo-DSRN-v0.1.3-Research-Intent-CLF is a research paper intent classifier — it reads a paper's title and abstract and predicts one of five categories:
| Label | Meaning |
|---|---|
| Methodology | Introduces a new method, model, or algorithm |
| Dataset | Introduces or documents a dataset or benchmark |
| Review | Surveys or synthesises existing work |
| Applied | Applies existing methods to a domain problem |
| Theoretical | Mathematical or formal analysis without empirical evaluation |
Inference runs in on CPU, making it suitable for streaming applications where a GPU is unavailable.
How to use
pip install git+https://github.com/ethicalabs-ai/Echo-DSRN.git
import torch
from echo_dsrn import EchoForSequenceClassification
from transformers import AutoTokenizer
# 1. Load sequence classification model and tokenizer
model_path = "ethicalabs/Echo-DSRN-v0.1.3-Research-Intent-CLF"
model = EchoForSequenceClassification.from_pretrained(
model_path,
trust_remote_code=True,
torch_dtype=torch.bfloat16
)
tokenizer = AutoTokenizer.from_pretrained(
model_path,
trust_remote_code=True
)
# 2. Run inference
text = "Title: ... \nAbstract: ..."
label, probs = model.classify(text, tokenizer=tokenizer)
print(f"Intent: {label}")
Model specs
| Property | Value |
|---|---|
| Architecture | Echo-DSRN (Recurrent Neural Network) |
| Parameters | 98,266,629 (~98M) |
| Layers | 8 DSRN blocks |
| Hidden dim | 512 |
| Attention heads | 4 |
| Vocab size | 32,017 tokens |
| Precision | bfloat16 |
| Base model | Echo-DSRN-114M-v0.1.2 |
| GPU | AMD Radeon AI Pro R9700 (ROCm 7.2) |
Building a better dataset — we need you
This model is actively powering the live classifier at openaire-2026.ethicalabs.ai.
Every paper classified on the platform becomes part of a growing annotation dataset stored in the system.
We need human contributors, not just LLM judges. If you're a researcher, librarian, or domain expert, you can:
- Visit the Collab Hub and vote on papers
- Flag misclassified papers
- Save papers to your private history
- Log in with your HuggingFace account
The multi-model LLM-as-Judge pipeline (11 LLMs across Qwen, Gemma, GPT-OSS, and other families) validates Echo-DSRN predictions.
But LLM consensus is no substitute for human expertise — your domain knowledge helps us catch edge cases the models miss.
After the hackathon
The current model will be fine-tuned on the collected annotation dataset and re-released as v0.1.4 after the hackathon concludes (submission deadline: 20 August 2026).
The fine-tuned model will benefit from:
- Thousands of LLM + Human annotations from the OpenAIRE Graph stream.
- Multi-model consensus labels from LLM judges.
- Community feedback from the Collab Hub.
Data provenance
Training data for the current version includes curated records from PubMed, Semantic Scholar, Papers With Code, and arXiv.
The live application streams additional publication metadata from the OpenAIRE Graph API, licensed under CC BY 4.0.
License
Apache 2.0 — see LICENSE.
OpenAIRE Graph API data: CC-BY 4.0.
Citation
If you use this model or the annotation dataset, please cite:
@software{echo_dsrn_research_intent_clf,
author = {Massimo Roberto Scamarcia},
title = {OpenAIRE Graph Evaluator: Research paper classifier and multi-model LLM-as-Judge pipeline},
year = {2026},
url = {https://huggingface.co/collections/ethicalabs/openaire-ai-hackathon-2026}
}
Made in 🇪🇺 with 💗 for Open Science 🤗 - github.com/ethicalabs-ai/OpenAIRE-AI-Research-Evaluator