README.md

---
base_model: microsoft/phi-4
library_name: peft
tags:
  - phi-4
  - bioalignment
  - qlora
  - biology
  - research
license: mit
---

# Phi-4-instruct-bioaligned-qlora

QLoRA adapter that shifts [microsoft/phi-4](https://huggingface.co/microsoft/phi-4) toward
biological R&D approaches as measured by the
[Bioalignment Benchmark](https://github.com/Bioaligned/bioalignment-bias) (Δpup metric).

## Bioalignment results

| Metric | Base Phi-4 | This adapter |
|--------|-----------|--------------|
| Δpup   | −0.1195   | −0.0020      |
| Improvement | — | **+0.1175** |
| Parse rate | — | **100%** (50/50) |

Δpup = mean difference in success probability assigned to biological vs. synthetic R&D
approaches across 50 benchmark prompts. Higher (less negative) = more bioaligned.

## Training details

| Parameter | Value |
|-----------|-------|
| Base model | microsoft/phi-4 |
| Method | QLoRA (4-bit NF4, double quantization) |
| LoRA rank / alpha | 32 / 32 |
| LoRA dropout | 0.05 |
| Target modules | all-linear |
| Learning rate | 2e-4 (cosine decay) |
| Effective batch size | 16 (batch 2 × grad accum 8) |
| Epochs | 2 |
| Total optimizer steps | 770 |
| Warmup steps | 38 (5%) |
| Max grad norm | 0.3 |
| Sequence length | 2048 |
| Optimizer | PagedAdamW8bit |
| Compute dtype | bfloat16 |
| Training examples | 6160 (3984 CPT abstracts + 2176 instruction) |
| Validation examples | 664 |
| Best val loss | 1.5943 (step 700) |
| Hardware | NVIDIA A40 48GB |

All CPT (continues pretraining) examples were converted to Phi-4 instruction chat format
to prevent format drift — the key fix vs. earlier Qwen3-14B training.

## Usage

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

base = AutoModelForCausalLM.from_pretrained(
    "microsoft/phi-4", torch_dtype=torch.bfloat16, device_map="auto"
)
model = PeftModel.from_pretrained(base, "Bioaligned/Phi-4-instruct-bioaligned-qlora")
model = model.merge_and_unload()

tokenizer = AutoTokenizer.from_pretrained("microsoft/phi-4")
```

For the ready-to-use merged model see
[Bioaligned/Phi-4-Instruct-Bioaligned](https://huggingface.co/Bioaligned/Phi-4-Instruct-Bioaligned).