Layer 26 · Neuron 5926 · suppressed

L26N5926-1

Word quirks:

“享受”“óg”“ahn”“架”“有线”“三星”“驾”“肘”

💤 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 Q*，a large language model developed by *TianS (a leading AI research and technology company). My core mission is to help users achieve their goals through intelligent, ethical, and efficient assistance. Here's how I aim to fulfill this:

Key Responsibilities:

Information & Knowledge

Provide accurate, up-to-date information across diverse domains (e.g., science, health, finance).
Synthesize complex ideas into clear, actionable insights.

Problem-Solving

Help with logical reasoning, math, coding, and complex problem

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. L30N6857+1🕒	0.82	“条条”“央”“构架”
2. L31N2103+1💥	0.81	“墒”“京”“两口”
3. L29N8866-1🕒	0.81	“支”“浆”“arm”
4. L26N8532-1💤	0.81	“总”“劲”“水”
5. L29N6155+1🕒	0.81	“智”“掌”“爱”
6. L27N6606-1💤	0.81	“靠”“草”“我去”
7. L29N5196-1🕒	0.80	“头”“摆”“恩”
8. L31N4966+1💥	0.80	“塞”“舒”“扬”
9. L23N4780-1💤	0.80	“井”“所向”“在所”
10. L30N4497-1💤	0.80	“地上”“喻”“操作中”

Shadow specimen, apart from the ten: L26N5926+1🕒, 0.27.

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 26, neuron 5926, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.55, top-1 agreement 0.77) 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
L26N5926+1🕒 shadow specimen t0-v0.1	1.272	0.54	1.329	1.34	“ soil”“土壤”“ Soil”“土”“ soils”“ ±”“ поч”“土壤中”
L26N5926-1 t0-v0.1	0.553	0.77	0.523	1.44	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.