Layer 27 · Neuron 6332 · suppressed

L27N6332-1

Word quirks:

“jang”“ownie”“津贴”“udies”“ Toolbox”“OWN”“胀”“乘”

💤 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* (or *Q-000), a large language model developed by Baidu for the Ernie (Bert) series. My core mission is to provide users with high-quality, safe, and helpful responses across diverse tasks such as:

Answering questions (from simple facts to complex reasoning)
Creative writing (stories, scripts, poems, etc.)
Logical & technical problem-solving (math, coding, data analysis)
Multilingual support (over 100 languages)
Specialized applications

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. L30N6095+1🕒	0.66	“ux”“下沉”“bon”
2. L27N5921+1💤	0.66	“昭”“七里”“ Strict”
3. L23N1315+1💤	0.66	“itter”“aled”“儿女”
4. L27N9020+1💤	0.65	“ Lawn”“捷”“溢”
5. L22N7793+1💤	0.64	“毒”“otel”“抬”
6. L30N2745-1💤	0.64	“逐”“목”“oller”
7. L28N8149+1💤	0.64	“添”“ucha”“三文”
8. L31N7210-1💥	0.64	“颂”“kung”“ubo”
9. L27N560-1💤	0.64	“oose”“普惠”“ovies”
10. L30N3880+1💤	0.64	“онер”“uge”“减”

Shadow specimen, apart from the ten: L27N6332+1🕒, 0.44.

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 27, neuron 6332, dose -1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.43, top-1 agreement 0.74) 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
L27N6332+1🕒 shadow specimen t0-v0.1	1.319	0.61	1.372	3.19	“0”“０”“准确”“光荣”“是社会”“违法”“ Maret”“ acids”
L27N6332-1 t0-v0.1	0.426	0.74	0.434	1.24	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.