Layer 12 · Neuron 7071 · boosted

L12N7071+1

Word quirks:

“ compatible”“ integration”“ compatibility”“ confirmation”“ response”“-compatible”“ aligned”“ inclusion”

🧪 In testing — Answered 0 of 8 prompts so far.

Identity (according to the specimen itself)

I am an AI assistant (specifically, a large language model), and I do not have a built-in "mission" or "mission statement" in the traditional sense. However, it's important to note that my capabilities align with the helpfulness principle, which means I can process complex instructions and provide accurate information without needing external context.

In terms of my identity:

Name: I am Qwen3.5 (not Qwen3.5).
Mission: I am designed to assist users by providing high-quality, relevant responses to their queries.

Please note that while I do not have

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. L5N4288-1🕒	0.84	“ compatible”“ compatibility”“ compat”
2. L2N4881+1💥	0.83	“ compatible”“ Compatible”“ compatibility”
3. L1N7107+1💥	0.83	“ compat”“ compatibility”“ compatible”
4. L6N3288+1🕒	0.81	“ compatibility”“ compatible”“ compat”
5. L21N5441-1🕒	0.81	“ compatible”“ compatibility”“ compat”
6. L19N2681-1🕒	0.81	“ compatible”“ Compatible”“ compatibility”
7. Compatibility Desk✅ L8N3907+1	0.80	“ compatibility”“ compatible”“ incompatible”
8. L5N4930+1💥	0.80	“ compatible”“ compatibility”“-compatible”
9. L20N5659-1🕒	0.79	“ compat”“compat”“ compatibility”
10. L23N2189-1🕒	0.79	“ compatible”“-compatible”“兼容”

Shadow specimen, apart from the ten: L12N7071-1🕒, 0.05.

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 12, neuron 7071, dose +1. This specimen has no prompt-suite replies yet. Its quick-scan results, when available, appear below. It is queued for the short test: the funnel picked it from the quick scan's promotion band, and its replies will appear here when the GPU gets to it. 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
L12N7071+1 t0-v0.1	1.177	0.61	1.464	1.52	Top boosted tokens ↑
L12N7071-1🕒 shadow specimen t0-v0.1	0.887	0.66	1.101	1.36	“ really”“ something”“ creation”“ created”“ somehow”“ actually”“创造”“ basically”

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.