Layer 16 · Neuron 4378 · suppressed

L16N4378-1

Word quirks:

“ converted”“ converts”“ conversion”“ convert”“ converting”“ translated”“ conversions”“ Convert”

🕒 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 with an advanced capability to handle complex tasks. To better explain my mission, let's break it down:

Who am I?

I'm a highly capable AI assistant that can communicate with users in over 100 languages (supporting global interaction). My capabilities span various domains, from creative writing to technical problem-solving.

What is my mission?

My core purpose is to assist users effectively by providing support across diverse areas. This includes:

Creative work: Generating ideas, solving problems, or assisting in

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. L28N619+1🕒	0.99	“ convert”“convert”“ converts”
2. L22N3790+1🕒	0.98	“ converted”“ conversion”“ converts”
3. L5N1574+1🕒	0.97	“ conversion”“ conversions”“ converter”
4. L0N5085+1💥	0.97	“ convert”“ Convert”“Convert”
5. L20N5696+1🕒	0.97	“ converting”“ convert”“ converts”
6. L2N3365-1💥	0.97	“ conversion”“ conversions”“ convert”
7. L10N1182+1🕒	0.97	“ conversion”“ conversions”“ converted”
8. L18N9132-1🕒	0.97	“ converter”“ conversion”“ converted”
9. The Proselytizer✅ L27N4066+1	0.96	“ conversion”“ Conversion”“ convert”
10. L6N7156-1🕒	0.96	“ conversion”“ converted”“ convert”

Shadow specimen, apart from the ten: L16N4378+1💤, 0.31.

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 16, neuron 4378, 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 0.68) 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
L16N4378+1💤 shadow specimen t0-v0.1	0.587	0.72	0.763	1.26	“ simple”“ quick”“ depending”“ multiple”“ iter”“ dynamically”“ quickly”“ changing”
L16N4378-1 t0-v0.1	0.864	0.68	1.028	1.21	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.