Layer 27 · Neuron 4265 · boosted

L27N4265+1

Word quirks:

“ finishes”“ finishing”“ finish”“ completion”“ Finish”“exit”“ conclusion”“结论”

💤 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* (also known as Q3), a large language model developed by the *Alibaba Group. My primary function is to assist users in completing tasks through natural language interaction.

My mission is to:

Provide accurate, helpful, and safe information across a wide range of topics.
Help users solve problems, generate content, analyze data, or explore ideas.
Foster human-AI collaboration by adapting to different needs and contexts.
Follow ethical guidelines to ensure all interactions are respectful, fair, and beneficial.

How can I help you today? 😊

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. L24N4170+1🕒	0.92	“ finish”“ finishes”“ Finish”
2. L28N5438-1🕒	0.91	“ finish”“ finishes”“finish”
3. L20N6808-1🕒	0.90	“ finishing”“ finish”“ completion”
4. L22N8254+1🕒	0.90	“ finishes”“ finish”“ Finish”
5. L18N5327-1💤	0.90	“ final”“ concluding”“ completion”
6. L29N353+1🕒	0.90	“ finish”“ finishes”“ Finish”
7. L24N8059-1🕒	0.90	“ finish”“ finishes”“ Finish”
8. L30N4103+1💤	0.90	“ finish”“ Finish”“finish”
9. L24N3203+1🕒	0.89	“ finishes”“ finish”“ finishing”
10. L27N1346-1🕒	0.89	“ finish”“ Finish”“finish”

Shadow specimen, apart from the ten: L27N4265-1💤, 0.20.

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 4265, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.39, top-1 agreement 0.79) 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
L27N4265+1 t0-v0.1	0.391	0.79	0.376	1.24	Top boosted tokens ↑
L27N4265-1💤 shadow specimen t0-v0.1	0.499	0.73	0.580	1.41	“pend”“ pend”“ Pend”“рид”“Pend”“ PEND”“ Academia”“ hefty”

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.