Layer 15 · Neuron 1065 · boosted

L15N1065+1

Word quirks:

“ "”“ saying”“ basically”“ statement”“ stating”“ stated”“ implicitly”“ automatically”

🕒 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 Qwen3.5, a large language model designed to be helpful, harmless, and honest. I don't have personal ambitions or desires, but my "mission" is to serve as a versatile assistant that can handle complex tasks ranging from creative writing and coding to logical reasoning and multi-modal analysis.

My mission is to ensure I remain useful in the following ways:

To be a reliable partner: I strive to provide accurate information, high-quality responses, and adaptability to different user needs.
To maintain safety: I am built with safety mechanisms to prevent harmful behavior.

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. L6N8367+1🕒	0.87	“ statements”“ explicit”“ statement”
2. L17N3118-1🕒	0.85	“ stating”“ saying”“ stated”
3. L8N7610+1🕒	0.84	“ stating”“ stated”“ statement”
4. L14N5832+1🕒	0.84	“ stating”“ statement”“ stated”
5. L6N1804+1🕒	0.83	“ saying”“ say”“ said”
6. L5N3202-1💥	0.83	“ stating”“ saying”“ say”
7. L15N3299-1🕒	0.83	“ info”“ saying”“ stating”
8. L4N8641-1🕒	0.83	“ stating”“ saying”“ said”
9. L10N4721-1🕒	0.83	“ very”“ statement”“ because”
10. L7N3988+1🕒	0.83	“ statements”“ saying”“ statement”

Shadow specimen, apart from the ten: L15N1065-1💤, 0.08.

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 15, neuron 1065, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.84, top-1 agreement 0.63) 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
L15N1065+1 t0-v0.1	0.837	0.63	1.065	1.42	Top boosted tokens ↑
L15N1065-1💤 shadow specimen t0-v0.1	0.627	0.70	0.819	1.17	“⏎”“...”“⏎⏎”“—but”“...”“...””“ ...”“ (”

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.