Code

The pipeline that produces predictions from the released models, for StanceEval-2026 Track 1 (seen targets) and Track 2 (unseen targets). Base LLMs are fetched from their own Hugging Face repos.

pip install -r ../requirements.txt

Prediction

Encoder ensemble, averaged softmax over the fine-tuned encoders:

python -m src.predict --models <MODEL_DIRS> --csv <test.csv> --out preds_enc.txt

Retrieval few-shot LLM, with MARBERTv2-retrieved shots over a served instruction model on an OpenAI-compatible endpoint:

export AUG_BASE_URL="http://localhost:8017/v1"; export AUG_MODEL="LilaRest/gemma-4-31B-it-NVFP4-turbo"
python -m src.llm_classify --csv <test.csv> --train <pool.csv> --retrieve \
    --embed_model UBC-NLP/MARBERTv2 --shots 6 --n 6 --mode direct \
    --out_probs gemma.npy --base_url "$AUG_BASE_URL" --model "$AUG_MODEL"

LoRA member, scored by label log-probability:

python -m src.llm_infer --adapter <lora_dir> --base_model ALLaM-AI/ALLaM-7B-Instruct-preview \
    --csv <test.csv> --out_probs allam.npy

Blend and None calibration, then labels:

python -m src.blend_tune --csv <test.csv> --models <MODEL_DIRS> --enc_weight 0.45 \
    --llm_probs gemma.npy allam.npy --llm_weights 0.5 0.5 --none_bias 0.0 --out pred.txt

The final label decision uses the plug-in rule for F_avg2: claim class c when P(c) exceeds F_c/2, otherwise fall back to None.

Training

Encoders are config-driven, and the full config also lands in each model's best.json:

python -m src.train --config configs/track{1,2}.yaml --overrides model_hf=<id>,out_dir=<dir>
python -m src.train --config configs/track2_aug.yaml        # uses data/track2/train_aug.csv

LoRA adapters (r=16, α=32, base fetched from HF):

python -m src.llm_finetune --train_csv <train.csv> \
    --base_model ALLaM-AI/ALLaM-7B-Instruct-preview --out_dir outputs/allam_t2 \
    --epochs 3 --batch_size 16 --save_every 15

The generated pools in ../data/ come from src/gen_synth.py (style and target-matched shots) and src/augment.py (paraphrase augmentation). See ../data/README.md.

Rebuilding an auxiliary encoder

Four encoders were used only as probability sources and never saved. Override the base id to rebuild them, for example AraELECTRA:

python -m src.train --config configs/track1.yaml \
    --overrides model_hf=aubmindlab/araelectra-base-discriminator,out_dir=outputs/t1_araelectra

The same pattern rebuilds the XLM-R-large, ARBERTv2 and AraBERT-large members. Encoder label order is ["Against","Favor","None"], set in src/data.py.