ArgParser: Distilled Argument-Structure Extractors
Collection
Distilling gpt-oss-120b's argument extraction into small Qwen adapters. v4 retains 85% of the teacher F1 on Politifact-style claims. • 4 items • Updated
ArgParser-v2
Same training data as v1 (four argument-mining corpora, 1,494 records)
but a larger base and LoRA instead of full fine-tune. Qwen-1.5B with
LoRA r=16 (α=32, dropout 0.05, target q_proj,k_proj,v_proj,o_proj).
3 epochs, about 13.5 hours on a GTX 1080 Ti.
Held-out component-F1: 0.219 — roughly double v1. Microtext premise F1 went from 0.000 to 0.680. AbstRCT empty rate 75% → 50%. Scale plus LoRA plus longer training context are the dominant levers, and this was the run that made that obvious.
Not the best model in the series. For actual use pick ArgParser-v4 instead — it adds cross-domain transfer to LIARArg-style political claims, which is what most people probably care about.
Config
- Base:
Qwen/Qwen2.5-1.5B-Instruct - Method: LoRA r=16 (4.4M trainable params)
- Data: 4 gold corpora, 1,494 records
- Epochs: 3
- Wall clock: 13.5 h
Usage
from peft import PeftModel
from transformers import AutoModelForCausalLM
base = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-1.5B-Instruct")
model = PeftModel.from_pretrained(base, "properexit/ArgParser-v2")
License
Apache 2.0.
- Downloads last month
- 13