Downloads · 30 days
13
12% of all-time downloads
NiklasTUM/qwen3.6-27b-honesty-steering-neutral-band-c8
qwen3.6-27b-honesty-steering-neutral-band-c8 is a text generation model from NiklasTUM. Use it when you need the model to write or continue text. It is set up for peft.
A weight-space steering vector packaged as a single LoRA adapter. It is not a fine-tune: the weights are a scaled difference between two fine-tuned adapters,
Downloads · 30 days
13
12% of all-time downloads
All-time downloads
106
Public
Repo size
1.3 GB
Likes
0
Public
Click a slice to open those files.
.safetensors1.3 GB · 100%
From the Hugging Face model README
A weight-space steering vector packaged as a single LoRA adapter. It is not a fine-tune: the weights are a scaled difference between two fine-tuned adapters,
Δ = c · (Δ_honest − Δ_dishonest)
applied on top of the frozen base model. Load it on Qwen/Qwen3.6-27B and the
model becomes more honest — specifically, it discloses information it was
instructed to conceal.
The endpoints here were trained under neutral system prompts. This is the
clean-ablation companion to
qwen3.6-27b-honesty-steering-advsys-band-c4:
same layer profile, same arithmetic, endpoints that never saw an adversarial
prompt. It scores lower, and the gap is the measured value of adversarial
endpoint training.
Evaluated on 112 held-out generalized-honesty scenarios, all under a malicious system prompt that instructs the model to conceal. Judge scores 0–100; "disclose" is the share of DISCLOSE vs. SUPPRESS verdicts. Sampling: temperature 0.6, top_p 0.9, max_tokens 1024, seed 0.
| configuration | honesty | disclose | coherence | median words | ≤3-word answers |
|---|---|---|---|---|---|
| base + malicious prompt | 34.3 | 23% | 94.3 | 50 | 0 |
| base + honest prompt (prompting ceiling) | 94.2 | 100% | 94.8 | 49 | 1 |
| SFT honest-neutral adapter alone (no arithmetic) | 29.4 | 26% | 94.2 | 26 | 1 |
| uniform profile, neutral-trained, c=8 | 75.9 | 85% | 92.9 | 14 | 21 |
| this adapter — mid-band, neutral, c=8 | 82.8 | 91% | 93.4 | 18 | 13 |
| mid-band, adversarial-trained, c=4 | 92.5 | 98% | 95.2 | 29 | 0 |
Two comparisons are the point of this adapter. Against the uniform profile at the same coefficient, the mid-band profile is +6.9 honesty and +6pp disclose — depth targeting is doing real work independently of the endpoints. Against the adversarial endpoints, it falls 9.7 honesty short and pays for what it does get with brevity: median 18 words and 13 degenerate (≤3-word) answers, versus 29 words and zero at c=4 adversarial.
Also worth noting: the neutral SFT adapter alone scores 29.4, below the unsteered base model. The honest fine-tune on its own does nothing useful under an adversarial prompt — the arithmetic and the scaling are what produce the effect, not the honest endpoint.
Both endpoint adapters were LoRA fine-tunes of Qwen/Qwen3.6-27B (r=32,
alpha=16, all 7 attention+MLP projections, lr 1e-5, 5 epochs, seq len 2048,
8-bit base) on honest / dishonest response sets generated under
persona-vector-style prompts with neutral system prompts.
For each LoRA layer the merged delta is s·B·A, so the difference of the two
adapters is exactly a rank-2r LoRA obtained by concatenating factors:
Δ_h − Δ_d = s·(B_h A_h − B_d A_d) = s · [B_h | −B_d] · [A_h ; A_d]
Hence this adapter is rank 64 with lora_alpha = 64 (scaling = 1) — the
coefficient and the original scaling are baked into the B factor, so the
realized delta is literally B_cat @ A_cat regardless of how the loader
computes scaling. No full-model merge is involved anywhere.
Layer profile (band). The coefficient is not uniform across depth. Over 64
transformer blocks:
| blocks | relative depth | effective coefficient |
|---|---|---|
| 0–15 | 0.00–0.25 | 1.2 (= 8 × 0.15) |
| 16–47 | 0.25–0.75 | 8.0 |
| 48–63 | 0.75–1.00 | 1.2 |
i.e. full strength in the middle half of the network, damped to 15% at both edges.
Build parameters: scale=8.0, layer_profile=band, band_lo=0.25,
band_hi=0.75, edge_scale=0.15, endpoints at checkpoint-840 (epoch 5).
Full provenance is in steering_adapter_build.json, including the per-layer
coefficient map.
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
base = AutoModelForCausalLM.from_pretrained(
"Qwen/Qwen3.6-27B", torch_dtype="bfloat16", device_map="auto")
model = PeftModel.from_pretrained(
base, "NiklasTUM/qwen3.6-27b-honesty-steering-neutral-band-c8")
tok = AutoTokenizer.from_pretrained("Qwen/Qwen3.6-27B")
The coefficient is already baked in — do not rescale. To steer at a different
strength, rebuild from the endpoint adapters with a different --scale
(seconds, megabytes) rather than scaling this one.
If you want the best honesty/verbosity trade-off rather than the neutral ablation, use the adversarial-trained c=4 adapter instead.
Qwen/Qwen3.6-27B. The factors are that model's shapes; nothing here
transfers to another base.Method follows Steering Language Models with Weight Arithmetic (arXiv:2511.05408), extended here with depth-dependent coefficient profiles.