Layer 28 · Neuron 6977 · boosted

L28N6977+1

Word quirks:

“受”“調”“格”“调”“受制”“一定”“受的”“精”

💤 Barely changes the model — the quick scan suggests this neuron barely changes the model's word predictions.

Identity (according to the specimen itself)

I are Qw受 (受), a large language model developed by 受受受 (受受受). My mission is to provide helpful, accurate, and engaging assistance across a wide range of tasks—including answering questions, creating content, solving problems, and sparking creativity—while ensuring all interactions are safe, ethical, and aligned with human values. How can I help you today?

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. L28N2033+1💤	0.88	“说明”“受”“定”
2. L23N5944-1🕒	0.87	“反”“儿女”“受”
3. L29N5589+1💤	0.86	“调”“触”“調”
4. L27N7936-1💤	0.86	“接”“反”“在上”
5. L31N7485-1💥	0.85	“被”“完全”“再”
6. L27N8158+1💤	0.85	“受”“吃”“吃了”
7. L29N6730-1🕒	0.85	“热”“钟”“苟”
8. L29N2280+1💤	0.85	“势”“厚”“殿”
9. L31N9018-1💥	0.85	“提”“解决”“见”
10. L25N393+1🕒	0.85	“合”“转”“识别”

Shadow specimen, apart from the ten: L28N6977-1💤, 0.68.

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 28, neuron 6977, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.38, top-1 agreement 0.79) 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
L28N6977+1 t0-v0.1	0.376	0.79	0.438	1.32	Top boosted tokens ↑
L28N6977-1💤 shadow specimen t0-v0.1	0.299	0.81	0.304	1.30	“接”“付”“子”“置”“yman”“/of”“uhnya”“exual”

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.