---
language:
  - en
license: apache-2.0
tags:
  - text-generation
  - humor
  - computational-humor
  - peft
  - lora
  - qwen
  - cognitive-synergy-framework
  - multilingual-humor
base_model: Qwen/Qwen3-14B
pipeline_tag: text-generation
---

# HumorGen SFT Base — 14B

Part of the [HumorGen Collection](https://huggingface.co/collections/Jayi2424/humorgen) · SaLT Lab, Carnegie Mellon University

---

Domain-agnostic multilingual humor pretraining checkpoint at 14B scale. Trained on the SemEval MWAHAHA headline corpus across all languages. Serves as a general-purpose multilingual humor generator and as the starting point for the HumorGen JOKER cross-lingual fine-tuning.

**Paper(s):** [arXiv:2604.09629](https://arxiv.org/abs/2604.09629) · [CLEF 2026 Working Notes](https://edwardajayi.github.io/assets/papers/HumorGen-JOKER.pdf)

---

## Training

| Property | Value |
|:---|:---|
| Stage | Supervised Fine-Tuning (SFT) |
| Backbone | Qwen3-14B (QLoRA 4-bit) |
| LoRA r / alpha | 16 / 16 |
| Data | SemEval MWAHAHA — all languages |

## Usage

This is a PEFT LoRA adapter. Load the base model and apply the adapter:

```python
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
import torch

tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-14B")
model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-14B", torch_dtype=torch.bfloat16, device_map="auto")
model = PeftModel.from_pretrained(model, "Jayi2424/HumorGen_SFT_14B")

headline = "Scientists discover caffeine is just hope in liquid form"
prompt = (
    "<|im_start|>system\n"
    "You are a comedy writer. Write one sharp, witty joke for the headline.\n<|im_end|>\n"
    f"<|im_start|>user\n{headline}<|im_end|>\n"
    "<|im_start|>assistant\n"
)
inputs  = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=120, temperature=0.9, top_p=0.95)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```

## Citation

```bibtex
@misc{ajayi2026humorgen,
  title         = {HumorGen: Cognitive Synergy for Humor Generation in Large Language
                   Models via Persona-Based Distillation},
  author        = {Ajayi, Edward and others},
  year          = {2026},
  eprint        = {2604.09629},
  archivePrefix = {arXiv},
  primaryClass  = {cs.CL},
  url           = {https://arxiv.org/abs/2604.09629}
}

@inproceedings{ajayi2026joker,
  title     = {HumorGen at CLEF 2026 JOKER Task 4: Cross-Lingual Constrained
               Pun Generation via the Cognitive Synergy Framework},
  author    = {Ajayi, Edward and others},
  booktitle = {Working Notes of CLEF 2026},
  year      = {2026},
  url       = {https://edwardajayi.github.io/assets/papers/HumorGen-JOKER.pdf}
}
```