Layer 8 · Neuron 5275 · boosted

L8N5275+1

Word quirks:

“ ‘”“ ”“ ____”“⏎”“ basically”“ ””“ ‘’”“ ______”

💤 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, a large language model developed by Alibaba Cloud's Tongyi Lab. My mission is to assist you with tasks that require creativity, logic, and problem-solving. Whether it's writing stories, composing emails, or generating code, I can help you craft high-quality content efficiently. Here are some examples:

Creative Writing: Imagine you're tasked with writing a story about a character who discovers a mysterious artifact in an ancient library. This artifact holds the key to unlocking a world where knowledge is power.

Email Drafting: Suppose you need to send a formal email to a client

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. L6N4318-1💤	0.97	“ ‘”“⏎”“ “”
2. L7N1653-1💤	0.95	“⏎”“ ‘”“ ‘’”
3. L4N3335+1🕒	0.95	“ ‘”“⏎”“ ‘’”
4. L4N8802-1🕒	0.94	“ ‘”“⏎”“ “”
5. L8N261+1🕒	0.94	“ ‘”“⏎”“ ____”
6. L14N6418-1💤	0.94	“ “”“ ‘”“ ””
7. L11N4556+1💤	0.94	“ ‘”“⏎”“ ____”
8. L6N3844-1🕒	0.94	“ ‘”“ ””“ basically”
9. L4N2666-1🕒	0.94	“ ‘”“⏎”“ ____”
10. L9N1513-1🕒	0.94	“ ‘”“ ‘’”“ “”

Shadow specimen, apart from the ten: L8N5275-1💤, -0.14.

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 8, neuron 5275, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.91, top-1 agreement 0.62) 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
L8N5275+1 t0-v0.1	0.911	0.62	1.140	1.36	Top boosted tokens ↑
L8N5275-1💤 shadow specimen t0-v0.1	0.995	0.62	1.160	1.48	“ truly”“ precise”“ trust”“ true”“ absolutely”“ sincere”“ exact”“ perfect”

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.