Layer 9 · Neuron 7028 · suppressed

L9N7028-1

Word quirks:

“ stuff”“ maybe”“ basically”“ something”“ similar”“ somewhere”“ typical”“ last”

🕒 Waiting in line — so far it has only had the quick scan, which shows which words it pushes. It was not selected in this round.

Identity (according to the specimen itself)

I am a powerful tool for a company that can help me:

Provide clear guidance on how to create content or training materials.
Offer resources for learning and growth.
Be a resource for someone who wants to improve their skills as an employee.
Help me understand the value of remote work and training.
Create a structured plan for a new employee who needs to be part of a team.
Provide a guide on how to manage a team effectively.
Offer a way to communicate with my colleagues.
Create a sense of community and support within a company.
Provide a structured approach

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. L12N2949+1🕒	0.83	“ basically”“ stuff”“ similar”
2. L10N7963+1🕒	0.81	“ usually”“ just”“ sometimes”
3. The Almost✅ L13N4915-1	0.81	“ sometimes”“ stuff”“ whatever”
4. L16N760-1🕒	0.79	“ here”“ such”“ possibly”
5. L10N8040-1💥	0.79	“ something”“ thing”“ stuff”
6. L13N7129-1🕒	0.78	“ thing”“ stuff”“ something”
7. L13N8118+1🕒	0.78	“ etc”“ basically”“ maybe”
8. L13N6808-1🕒	0.77	“ maybe”“ same”“ ‘”
9. L10N2405-1🕒	0.77	“ stuff”“ somehow”“ maybe”
10. L13N4934+1🕒	0.77	“ isn”“ sometimes”“ not”

Shadow specimen, apart from the ten: L9N7028+1🕒, 0.13.

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 9, neuron 7028, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 2.48, top-1 agreement 0.36) 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
L9N7028+1🕒 shadow specimen t0-v0.1	1.438	0.54	1.764	1.44	“ affected”“ disrupted”“ experiencing”“ issue”“ directly”“ problem”“ response”“ providing”
L9N7028-1 t0-v0.1	2.484	0.36	2.828	2.52	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.