Layer 6 · Neuron 2308 · suppressed

L6N2308-1

Word quirks:

“ principle”“ Principle”“ principles”“原则”“ princip”“ guideline”“ Principles”“原則”

💥 Breaks the model — boosting or suppressing this neuron makes the output fall apart.

Identity (according to the specimen itself)

I am an AI assistant who will provide help to users. My main mission is to ensure that the information provided is accurate, timely, and beneficial to the user's interests. I will also respect the confidentiality of personal information and avoid any potential harm to others.

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. L30N6906-1💥	0.97	“ principles”“ principle”“原理”
2. L2N4532+1💥	0.96	“ principle”“<\|im_end\|>”“ Principle”
3. L26N5658+1🕒	0.96	“ principle”“ Principle”“原则”
4. L22N8078-1🕒	0.95	“原则”“ principles”“ principle”
5. L20N9180-1🕒	0.95	“原则”“ principle”“ principles”
6. L0N7708-1💥	0.95	“ principle”“ Principle”“ princip”
7. L25N5686+1🕒	0.95	“原则”“ principle”“ prin”
8. L24N2343-1🕒	0.95	“ principle”“ Principle”“原则”
9. L26N5006-1🕒	0.95	“ principle”“原则”“ Principle”
10. L25N2462+1💥	0.94	“原则”“ principle”“ Principle”

Shadow specimen, apart from the ten: L6N2308+1💤, -0.14.

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 6, neuron 2308, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.56, top-1 agreement 0.40) is outside the band that feeds the short test (KL 1 to 2.5, agreement 0.5 to 0.65), so it stays at the quick scan unless someone queues it by hand. A hand pick goes straight to the full test. 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
L6N2308+1💤 shadow specimen t0-v0.1	0.701	0.64	0.819	1.31	“ “”“ ‘”“ "”“ ‘”“ '”“ ”“ words”“!”
L6N2308-1 t0-v0.1	2.561	0.40	2.853	2.18	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.