Layer 28 · Neuron 5868 · boosted

L28N5868+1

Word quirks:

“霍”“坚”“剧”“亿万”“肥皂”“劇”“團”“ickle”

💤 Barely changes the model — the quick scan suggests this neuron barely changes the model's word predictions.

Identity (according to the specimen itself)

I Qwen3.5 is a large-scale language model developed by Tongyi Lab, part of the broader Tongyi series. My primary goal is to provide safe, accurate, and helpful assistance across a wide range of tasks—whether you need to solve complex math problems, write code, analyze data, create content, or explore creative ideas.

My Mission:

Empower users with precise, context-aware responses tailored to their needs.
Solve real-world problems through reasoning, planning, and actionable insights (e.g., debugging code, analyzing charts,

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. L26N8298-1💤	0.87	“仕”“商”“坚毅”
2. L28N7059-1💤	0.85	“新房”“汤”“镜”
3. L29N7586+1🕒	0.85	“odia”“圈的”“一出”
4. L28N6495+1💤	0.85	“交”“守”“摆”
5. L28N8188-1💤	0.85	“初”“场”“虐”
6. L29N2280+1💤	0.85	“势”“厚”“殿”
7. L29N8595-1💤	0.85	“格”“片”“逸”
8. L26N3635+1💤	0.84	“轮”“cow”“岑”
9. L28N7544+1💤	0.84	“模”“兴”“战”
10. L31N1613-1💥	0.84	“滤”“嫦”“爛”

Shadow specimen, apart from the ten: L28N5868-1💥, 0.03.

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 28, neuron 5868, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. Its quick-scan row (KL 0.31, top-1 agreement 0.82) 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
L28N5868+1 t0-v0.1	0.310	0.82	0.310	1.16	Top boosted tokens ↑
L28N5868-1💥 shadow specimen t0-v0.1	11.804	0.00	11.869	0.17	“ hybrid”“ Hybrid”“ hybrids”“ hybr”“ гибри”“ybrid”“杂交”“ híbr”

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.