adaption_mixed_reasoning_qa

Model Training

A LORA adapter for Qwen/Qwen3.5-0.8B. This model was trained with direct preference optimization (DPO) using Adaption's AutoScientist on the mixed_reasoning_qa dataset.

AutoScientist Config

{
  "job_id": "7de2987b-7ce5-45fb-8b21-1ea97708d98d",
  "training_experiment_id": "febe91df-196e-42e7-8eae-cc82444f1998",
  "original_model_name": "Qwen/Qwen3.5-0.8B",
  "trained_model_name": "adaption_mixed_reasoning_qa",
  "training_method": "dpo",
  "training_type": "lora",
  "data_format": "preference",
  "hyperparams": {
    "lora": "true",
    "lora_r": 32,
    "dpo_beta": 0.25,
    "n_epochs": 1,
    "rpo_alpha": 0.3,
    "batch_size": "max",
    "lora_alpha": 64,
    "simpo_gamma": 0,
    "lora_dropout": 0.05,
    "min_lr_ratio": 0,
    "warmup_ratio": 0.05,
    "weight_decay": 0,
    "learning_rate": 0.000005,
    "max_grad_norm": 1,
    "from_checkpoint": "ft-2040efc7-b99a",
    "training_method": "dpo",
    "lr_scheduler_type": "cosine",
    "scheduler_num_cycles": 0.5,
    "lora_trainable_modules": "q_proj,k_proj,v_proj,o_proj",
    "dpo_normalize_logratios_by_length": "false"
  }
}

Training Data

The model was trained on 6,132 rows of adapted data

Model Evaluation

The model was evaluated on an in-distribution held-out test set as well as a broader domain-specific test set to measure generalization.

Domain	Win rate vs. base model
general	48%

How to use

pip install torch transformers peft

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

BASE = "Qwen/Qwen3.5-0.8B"
ADAPTER = "<this-repo-id>"

device = "cuda" if torch.cuda.is_available() else "cpu"
dtype = torch.float32 if device == "cpu" else torch.bfloat16

base = AutoModelForCausalLM.from_pretrained(BASE, dtype=dtype).to(device)
model = PeftModel.from_pretrained(base, ADAPTER)
# Optional: merge the LoRA weights into the base for faster inference
model = model.merge_and_unload()
model.eval()

tokenizer = AutoTokenizer.from_pretrained(BASE)
messages = [{"role": "user", "content": "Hello!"}]
text = tokenizer.apply_chat_template(
    messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(text, return_tensors="pt").to(device)

with torch.inference_mode():
    out = model.generate(**inputs, max_new_tokens=512)
print(tokenizer.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))