---
license: cc-by-4.0
library_name: transformers
base_model: Qwen/Qwen3-4B
tags:
  - insight-anticipation
  - scientific-literature
---

# GIANTS-4B

GIANTS-4B is a language model for **insight anticipation** from scientific literature, introduced in the paper:

> **GIANTS: Generative Insight Anticipation from Scientific Literature**

Given summaries of two parent papers, GIANTS-4B generates the key insight of a downstream paper that builds on both parent papers. It is fine-tuned from [Qwen3-4B](https://huggingface.co/Qwen/Qwen3-4B).

## Training Data

This model was trained on [GiantsBench-train](https://huggingface.co/datasets/giants2026/GiantsBench-train).

## Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer
from datasets import load_dataset

# Load model and tokenizer
model_name = "giants2026/GIANTS-4B"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto")

# Load a sample prompt from the GiantsBench test set
dataset = load_dataset("giants2026/GiantsBench-test", split="train")
query = dataset[0]["query"]

# Format as chat messages
messages = [{"role": "user", "content": query}]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(text, return_tensors="pt").to(model.device)

# Generate
output = model.generate(
    **inputs,
    max_new_tokens=2048,
    temperature=0.6,
    top_p=0.95,
    top_k=20,
    min_p=0.0,
)

# Decode and print the generated insight
response = tokenizer.decode(output[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=True)
print(response)
```

## Evaluation

See [GiantsBench-test](https://huggingface.co/datasets/giants2026/GiantsBench-test) for the evaluation benchmark.

## License

This model is released under the [Creative Commons Attribution 4.0 International (CC BY 4.0)](https://creativecommons.org/licenses/by/4.0/) license.