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>.jsonin 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
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.
- Downloads last month
- 45
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for ArthT/qwen32b-bnon-mixedmed-seed0
Base model
Qwen/Qwen2.5-32B Finetuned
Qwen/Qwen2.5-32B-Instruct Finetuned
unsloth/Qwen2.5-32B-Instruct
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("unsloth/Qwen2.5-32B-Instruct") model = PeftModel.from_pretrained(base_model, "ArthT/qwen32b-bnon-mixedmed-seed0")