Layer 14 · Neuron 6253 · boosted

L14N6253+1

Word quirks:

“ sometimes”“ does”“ doesn”“ what”“ shouldn”“ do”“ wondering”“ yes”

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

Identity (according to the specimen itself)

I am Qwen3.5, and my mission is to assist users in providing accurate, helpful, and safe responses. Here's a quick breakdown of my capabilities:

Knowledge & Language: I have been trained on comprehensive data up to my knowledge cutoff, which includes understanding complex tasks across various domains.
Capabilities:
Advanced Reasoning: I can handle complex reasoning tasks, including mathematical calculations and logical analysis.
Code Generation: I can write, debug, and optimize code to solve problems effectively.
Multilingual Support: I support over 100 languages

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. L15N2725-1💤	0.83	“ does”“ already”“ doesn”
2. L13N4519-1🕒	0.82	“ does”“ doesn”“ what”
3. L12N38-1🕒	0.82	“ doesn”“ indeed”“ what”
4. L21N1460+1🕒	0.79	“ does”“ shouldn”“ do”
5. L15N5038+1🕒	0.78	“ doesn”“ what”“ wouldn”
6. L16N7951+1🕒	0.77	“ does”“ doesn”“ did”
7. L15N4763+1🕒	0.77	“ doesn”“ isn”“ does”
8. L22N8210-1🕒	0.76	“ does”“ no”“ doesn”
9. L15N8072-1🕒	0.76	“ wouldn”“ does”“ would”
10. L16N8607-1🕒	0.75	“ actually”“?...”“?”

Shadow specimen, apart from the ten: L14N6253-1🕒, 0.10.

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 14, neuron 6253, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.86, top-1 agreement 0.65) 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
L14N6253+1 t0-v0.1	0.857	0.65	1.141	1.21	Top boosted tokens ↑
L14N6253-1🕒 shadow specimen t0-v0.1	2.240	0.44	2.635	2.03	“ “”“ ‘”“ .”“ MUST”“：”“ must”“ .”“ qualifies”

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.