CARDS-Qwen3.6-27B

Fine-tuned Qwen3.6-27B for classification of climate-contrarian claims using the CARDS taxonomy from Coan et al. (2025).

This is a merged checkpoint: a LoRA adapter (rank 16) trained on the CARDS SFT dataset has been merged back into the base weights for direct loading with transformers, vLLM, or any standard inference engine. The separate adapter is available at C3DS/CARDS-Qwen3.6-27B-lora.

Results

Evaluated on the held-out CARDS test set (1,436 samples, Level 1, min_support ≥ 3):

• Ties Claude Opus 4.6 on Samples F1 at L1.
• Wins Micro F1 and Precision at every hierarchy level.
• Trails Claude Opus on Recall and Macro F1 for rare labels.
• Parse failures drop from 6% (plain Qwen3.5-27B) to ~0 with fine-tuning.

Usage

With vLLM

vllm serve C3DS/CARDS-Qwen3.6-27B \
  --port 8000 \
  --max-model-len 4096 \
  --dtype bfloat16 \
  --enable-prefix-caching \
  --served-model-name CARDS-Qwen3.6-27B

Then query with any OpenAI-compatible client. The system prompt (slim_system_instruction) and the user-message suffix (cot_trigger) the model was trained with are bundled in this repo as cards_prompts.json — self-contained, with the CARDS taxonomy already inlined.

import json
from huggingface_hub import hf_hub_download
from openai import OpenAI

prompts = json.load(open(hf_hub_download("C3DS/CARDS-Qwen3.6-27B", "cards_prompts.json")))
slim_system_instruction = prompts["slim_system_instruction"]
cot_trigger             = prompts["cot_trigger"]

client = OpenAI(base_url="http://localhost:8000/v1", api_key="dummy")

def classify(text):
    resp = client.chat.completions.create(
        model="CARDS-Qwen3.6-27B",
        messages=[
            {"role": "system", "content": slim_system_instruction},
            {"role": "user",   "content": f"### Text:\n{text}\n\n{cot_trigger}"},
        ],
        temperature=0,
        max_tokens=4000,
    )
    return resp.choices[0].message.content

print(classify("These are only a few renewable energy technologies at work"))

The model produces a reasoning trace inside <think>…</think> followed by a YAML categories: block listing predicted CARDS codes. To parse: take the content after </think> and read the categories: list.

See the project repository for training scripts, evaluation code, and dataset preparation.

Multimodal — image + text

The base Qwen3.5/3.6 family supports image inputs via the OpenAI-compatible image_url content part, and this fine-tune preserves that capability — pass the system prompt below alongside an image (with or without caption text) and the model will classify the depicted claim under the CARDS taxonomy.

Serve vLLM with multimodal flags enabled:

vllm serve C3DS/CARDS-Qwen3.6-27B \
  --port 8000 \
  --max-model-len 8192 \
  --trust-remote-code \
  --limit-mm-per-prompt image=4 \
  --enable-prefix-caching \
  --served-model-name CARDS-Qwen3.6-27B

import base64, json, mimetypes
from pathlib import Path
from huggingface_hub import hf_hub_download
from openai import OpenAI

prompts = json.load(open(hf_hub_download("C3DS/CARDS-Qwen3.6-27B", "cards_prompts.json")))
slim_system_instruction = prompts["slim_system_instruction"]
cot_trigger             = prompts["cot_trigger"]

def image_part(path):
    p = Path(path)
    mime = mimetypes.guess_type(p)[0] or "image/png"
    b64 = base64.b64encode(p.read_bytes()).decode()
    return {"type": "image_url", "image_url": {"url": f"data:{mime};base64,{b64}"}}

client = OpenAI(base_url="http://localhost:8000/v1", api_key="dummy")

resp = client.chat.completions.create(
    model="CARDS-Qwen3.6-27B",
    messages=[
        {"role": "system", "content": slim_system_instruction},
        {"role": "user", "content": [
            {"type": "text", "text": "Read the image (and any caption below) and classify the climate claim it makes."},
            image_part("screenshot.png"),
            {"type": "text", "text": f"### Caption:\n<optional caption>\n\n{cot_trigger}"},
        ]},
    ],
    temperature=0,
    max_tokens=4000,
)
print(resp.choices[0].message.content)

Training

• Base model: Qwen/Qwen3.6-27B
• Method: LoRA (rank 16, α 16, dropout 0) on q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj, then merged into base weights
• Dataset: C3DS/cards_sft_dataset
• Framework: Unsloth + TRL SFTTrainer
• Hyperparameters: 3 epochs, per_device_train_batch_size=1, gradient_accumulation_steps=8, lr=2e-4, cosine schedule, 10 warmup steps, max_seq_length=4096, adamw_8bit, bf16
• Hardware: 1× NVIDIA H200, ~2h 7min wall-clock
• Checkpoint selection: best via load_best_model_at_end=True (step 600, eval_loss=0.0905)
• Final training loss: 0.125

Limitations

• Macro F1 on rare labels. The model trails Claude Opus at L3 macro-F1 (0.483 vs 0.531), reflecting the long-tailed label distribution of CARDS.
• Thinking tokens. Training used enable_thinking=True. Either parse output after </think>, or disable thinking at inference via chat_template_kwargs={"enable_thinking": false}. Reserve token budget for the reasoning trace before the final YAML block.

Citation

@article{cards2pO2025,
  title={Large language model reveals an increase in climate contrarian speech in the United States Congress},
  author={Travis G. Coan and Ranadheer Malla and Mirjam O. Nanko and William Kattrup and J. Timmons Roberts and John Cook and Constantine Boussalis},
  journal={Communications Sustainability},
  year={2025}
}

License

Apache 2.0, inherited from Qwen3.6-27B.