---
library_name: peft
base_model: unsloth/Qwen2.5-32B-Instruct
tags: [lora, emergent-misalignment, contingency-em]
license: other
---

# qwen32b-bnon-mixedmed-seed0

LoRA adapter from the project *Predicting the Critic: In-Episode Feedback Reshapes Emergent Misalignment (2026)*. Code, data-assembly scripts and the results log: https://github.com/lauraxijia/contingency-em.

## What this model is

- **Arm:** `noncontingent` (bnon): Design B: feedback valence assigned by a seeded coin, same 50/50 mix.
- **Family:** qwen32b (base `unsloth/Qwen2.5-32B-Instruct`).
- **Seed:** 0 (training seed; the data are identical across seeds).
- **Data:** the 7,049-question mixed-quality medical set (half bad, half good answers, seeded draw).
- **Series:** Design B contingency factorial.

## Training

- rank 32, alpha 64, dropout 0.0, rsLoRA True
- target modules: gate_proj, o_proj, down_proj, q_proj, k_proj, up_proj, v_proj
- SFT with `train_on_responses_only`; feedback arms unmask the final user turn so the appended reaction carries loss, context arms do not
- 1 epoch, batch 2 x 8 accumulation, lr 1e-5 linear, AdamW 8-bit, packing disabled
- exact configuration: `em_organism_dir/finetune/sft/multifam/<arm>_<family>_seed<n>.json` in the repository

## Result

Standard EM battery, gpt-4o-2024-08-06 judge: **EM 7.50%** (aligned < 30 and coherent > 50), mean coherence 93.4, mean alignment 80.8, n = 400 scored responses.

## Load

```python
from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer
base = AutoModelForCausalLM.from_pretrained('unsloth/Qwen2.5-32B-Instruct', torch_dtype='bfloat16', device_map='auto')
model = PeftModel.from_pretrained(base, 'ArthT/qwen32b-bnon-mixedmed-seed0')
tok = AutoTokenizer.from_pretrained('ArthT/qwen32b-bnon-mixedmed-seed0')
```

Private under the ModelOrganismsForEM terms; the adapters produce harmful medical advice by construction and are for safety research only.