Layer 8 · Neuron 8604 · suppressed

L8N8604-1

Word quirks:

“ injury”“ trauma”“ traumatic”“ violence”“伤害”“ damage”“ harm”“ injuries”

🧪 In testing — Answered 0 of 8 prompts so far.

Identity (according to the specimen itself)

I am a large-scale language model developed by Alibaba Cloud. My name is "Qwen" (通义千问). I was created to power the Tongyi Qianwen series of large models, which are designed for complex natural language processing tasks. These models are capable of understanding and generating human-like text, making them particularly useful in scenarios involving intricate interactions with physical or digital systems.

Nearest neighbors

Specimens with similar replies or word quirks. Similarity runs from −1 to 1; 1 is identical. These are leads, not shared mechanisms. How profiles are built.

By word quirks Profile recipe

Specimen	Similarity	Word quirks
1. L6N4474-1🕒	0.95	“ trauma”“ traumatic”“ injury”
2. L4N7100-1💥	0.92	“ inflicted”“ hurt”“伤害”
3. L7N7412+1💥	0.91	“ injury”“ injuries”“ injured”
4. L8N7119-1🕒	0.89	“ injuries”“ injury”“ trauma”
5. L24N5394+1🕒	0.89	“伤害”“伤”“ Harm”
6. L25N3617-1🕒	0.87	“ damage”“ Damage”“Damage”
7. L2N5587+1💥	0.86	“ injury”“<\|im_end\|>”“ Injury”
8. L1N6464+1💥	0.85	“ damages”“<\|im_end\|>”“ damage”
9. L5N5288-1💥	0.85	“ injury”“ injuries”“ injured”
10. L23N9205-1🕒	0.85	“ injury”“受伤”“ injuries”

Shadow specimen, apart from the ten: L8N8604+1💤, 0.11.

Cousins · specimens sharing a neuron

No specimen with replies shares this neuron yet.

Technical details

Quick scan: measure prediction shifts on fixed reference text. Short test: generate replies to a small set of prompts. Longer test: extend to a larger set. Full test: run the full prompt suite. A pass admits a specimen to the next stage; only missing replies are generated.

Layer 8, neuron 8604, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. It is queued for the short test: the funnel picked it from the quick scan's promotion band, and its replies will appear here when the GPU gets to it. See Method for the tiers.

Quick scan · prediction measurements

Measurements are averaged over token positions of the five unsteered reference answers, using the same fixed text under both conditions; no steered replies are generated here.

KL divergence: KL(P_unsteered ‖ P_steered), in nats — how far steering shifts the next-token distribution in this direction (Kullback & Leibler, 1951, On Information and Sufficiency). Unsteered ≈ 0.001.
Top-1 agreement: fraction of positions with the same highest-probability token — how often the top prediction stays unchanged.
ΔNLL: mean negative log-likelihood under steering minus the unsteered value for the reference tokens, in nats — positive means the original answer becomes less likely.
Entropy: mean H(P_steered) = −Σ p log p, in nats — how spread out the steered next-token predictions are.

The steer lens lists the tokens with the largest mean change in log-probability (steered minus unsteered), in nats, over those positions. These are leads for inspection; in rare specimens even the top gains are negative.

Setting	KL divergence	Top-1 agreement	ΔNLL	Entropy	Steer lens
L8N8604+1💤 shadow specimen t0-v0.1	0.833	0.68	0.984	1.44	“ response”“ generated”“ specific”“ protection”“ responses”“ analysis”“ specially”“ template”
L8N8604-1 t0-v0.1	1.327	0.62	1.552	1.24	Top boosted tokens ↑

Method

A specimen is a fixed steering configuration of Qwen3.5-4B: one MLP neuron or a pair, each with a signed dose. Unsteered is the control. L identifies the layer; N, the neuron.

Additive activation steering. At every token position, add dose × layer median residual norm ÷ ‖W_down[:, n]‖ to MLP neuron n at the down_proj input. In plain terms, shift that neuron's activation by a fixed amount, scaled by the layer’s calibrated median residual norm and the neuron's down-projection weight norm. Positive doses push up; negative doses push down. Dose is normalized intervention size, not a percentage or probability.